Federal Reserve Bank of New York
Economic September 2008 September Volume 14 Number 2 14 Number Volume Policy Review
Special Issue: The Economics of Payments ECONOMIC POLICY REVIEW
EDITOR Kenneth D. Garbade
COEDITORS Mary Amiti Adam B. Ashcraft Robert W. Rich Asani Sarkar
EDITORIAL STAF F Valerie LaPorte Mike De Mott Michelle Bailer Karen Carter
PRODUCTION STAFF Carol Perlmutter David Rosenberg Jane Urry
The Economic Policy Review is published by the Research and Statistics Group of the Federal Reserve Bank of New York. Articles undergo a comprehensive refereeing process prior to their acceptance in the Review. The views expressed are those of the individual authors and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System. www.newyorkfed.org/research Federal Reserve Bank of New York Economic Policy Review
September 2008 Volume 14 Number 2
Special Issue: The Economics of Payments
Contents
3Introduction James McAndrews
Theoretical Models of Money and Payments Articles:
7 Intraday Liquidity Management: A Tale of Games Banks Play Morten L. Bech
Over the last few decades, most central banks, concerned about settlement risks inherent in payment netting systems, have implemented real-time gross settlement (RTGS) systems. Although RTGS systems can significantly reduce settlement risk, they require greater liquidity to smooth nonsynchronized payment flows. Thus, central banks typically provide intraday credit to member banks, either as collateralized credit or priced credit. Because intraday credit is costly for banks, how intraday liquidity is managed has become a competitive parameter in commercial banking and a policy concern of central banks. This article uses a game-theoretical framework to analyze the intraday liquidity management behavior of banks in an RTGS setting. The games played by banks depend on the intraday credit policy of the central bank and encompass two well-known paradigms in game theory: “the prisoner's dilemma” and “the stag hunt.” The former strategy arises in a collateralized credit regime, where banks have an incentive to delay payments if intraday credit is expensive, an outcome that is socially inefficient. The latter strategy occurs in a priced credit regime, where postponement of payments can be socially efficient under certain circumstances. The author also discusses how several extensions of the framework affect the results, such as settlement risk, incomplete information, heterogeneity, and repeated play. 25 An Economic Analysis of Liquidity-Saving Mechanisms Antoine Martin and James McAndrews
A recent innovation in large-value payments systems has been the design and implementation of liquidity-saving mechanisms (LSMs), tools used in conjunction with real-time gross settlement (RTGS) systems. LSMs give system participants, such as banks, an option not offered by RTGS alone: they can queue their outgoing payments. Queued payments are released if some prespecified event occurs. LSMs can reduce the amount of central bank balances necessary to operate a payments system as well as quicken settlement. This article analyzes the performance of RTGS systems with and without the addition of an LSM. The authors find that, in terms of settling payments early, these mechanisms typically outperform pure RTGS systems. However, there are times when RTGS systems can be preferable to LSMs, such as when many banks that send payments early in RTGS choose to queue their payments when an LSM is available. The authors also show that the design of a liquidity-saving mechanism has important implications for the welfare of system participants, even in the absence of payment netting. In particular, the parameters specified determine whether the addition of an LSM increases or decreases welfare.
41 Divorcing Money from Monetary Policy Todd Keister, Antoine Martin, and James McAndrews
Many central banks implement monetary policy in a way that maintains a tight link between the stock of money and the short-term interest rate. In particular, their implementation procedures require that the supply of reserve balances be set precisely in order to implement the target interest rate. Because bank reserves play other key roles in the economy, this link can create tensions with other important objectives, especially in times of acute market stress. This article considers an alternative approach to monetary policy implementation—known as a “floor system”—that can reduce or even eliminate these tensions. The authors explain how this approach, in which the central bank pays interest on reserves at the target interest rate, “divorces” the supply of money from the conduct of monetary policy. The quantity of bank reserves can then be set according to the payment or liquidity needs of financial markets. By removing the opportunity cost of holding reserves, the floor system also encourages the efficient allocation of resources in the economy. Empirical Analyses of Trends in Large-Value Payments Articles:
59 Global Trends in Large-Value Payments Morten L. Bech, Christine Preisig, and Kimmo Soramäki
Globalization and technological innovation are two major forces affecting the financial system and its infrastructure. Perhaps nowhere are these trends more apparent than in the internationalization and automation of payments. While the effects of globalization and technological innovation are most obvious on retail payments, the influence is equally impressive on wholesale, or interbank, payments. Given the importance of payments and settlement systems to the smooth operation and resiliency of the financial system, it is important to understand the potential consequences of these developments. This article presents ten major long-range trends in the settlement of large-value payments worldwide. The trends are driven by technological innovation, structural changes in banking, and the evolution of central bank policies. The authors observe that banks, to balance risks and costs more effectively, are increasingly making large-value payments in real-time systems with advanced liquidity-management and liquidity-saving mechanisms. Moreover, banks are settling a larger number of foreign currencies directly in their home country by using offshore systems and settling a greater number of foreign exchange transactions in Continuous Linked Settlement Bank or through payment-versus-payment mechanisms in other systems. The study also shows that the service level of systems is improving, through enhancements such as longer operating hours and standardized risk management practices that adhere to common standards, while transaction fees are decreasing. Payments settled in large-value payments systems are more numerous, but on average of smaller value. Furthermore, the overall nominal total value of large-value payments is increasing, although the real value is declining.
83 Changes in the Timing Distribution of Fedwire Funds Transfers Olivier Armantier, Jeffrey Arnold, and James McAndrews
The Federal Reserve’s Fedwire funds transfer service—the biggest large-value payments system in the United States—has long displayed a peak of activity in the late afternoon. Theory suggests that the concentration of late-afternoon Fedwire activity reflects coordination among participating banks to reduce liquidity costs, delay costs, and credit risk; as these costs and risk change over time, payment timing most likely will be affected. This article seeks to quantify how the changing environment in which Fedwire operates has affected the timing of payment value transferred within the system between 1998 and 2006. It finds that the peak of the timing distribution has become more concentrated, has shifted to later in the day, and has actually divided into two peaks. The authors suggest that these trends can be explained by a rise in the value of payments transferred over Fedwire, the settlement patterns of the private settlement institutions that use the system, and an increase in industry concentration. Although the study’s results provide no specific evidence of heightened operational risk attributable to activity occurring later in the day, they point to a high level of interaction between Fedwire and private settlement institutions. 113 The Timing and Funding of CHAPS Sterling Payments Christopher Becher, Marco Galbiati, and Merxe Tudela
Real-time gross settlement (RTGS) systems such as CHAPS Sterling require large amounts of liquidity to support payment activity. To meet their liquidity needs, RTGS participants borrow from the central bank or rely on incoming payments from other participants. Both options can prove costly—the latter in particular if participants delay outgoing payments until incoming ones arrive. This article presents an empirical analysis of the timing and funding of payments in CHAPS. The authors seek to identify the factors driving the intraday profile of payment activity and the extent to which incoming funds are used as a funding source, a process known as liquidity recycling. They show that the level of liquidity recycling in CHAPS is high and stable throughout the day, and attribute this result to several features of the system. First, the settlement of time- critical payments provides liquidity to the system early in the settlement day; this liquidity can be recycled for the funding of less urgent payments. Second, CHAPS throughput guidelines provide a centralized coordination mechanism, in effect limiting any tendency toward payment delay. Third, the relatively small direct membership of CHAPS facilitates coordination between members, for example, through the use of bilateral net sender limits. Coordination encourages banks to maintain a relatively constant flux of payments throughout the day. The authors also argue that the high level of recycling helps to reduce liquidity risk, and that the relatively smooth intraday distribution of payments serves to mitigate operational risk associated with highly concentrated payment activity. They note, however, that the benefits of liquidity recycling are not evenly distributed between members of CHAPS.
Risk Management in Payments Systems Articles:
137 Understanding Risk Management in Emerging Retail Payments Michele Braun, James McAndrews, William Roberds, and Richard Sullivan
New technologies used in payment methods can reduce risk, but they can also lead to new risks. Emerging retail payments are prone to operational and fraud risks, especially security breaches and potential use in illicit transactions. This article describes an economic framework for understanding risk control in retail payments. Risk control is a special type of good because it can protect one payment participant without diminishing the protection of other participants. As a result, the authors’ economic framework emphasizes risk containment, primarily through the establishment and enforcement of risk management policies. Application of the framework to three types of emerging payments suggests that a payments system can successfully manage risk if it quickly recognizes problems, encourages commitment from all participants to control risk, and uses an appropriate mix of market and public policy mechanisms to align risk management incentives. The authors conclude that providers of emerging payment methods must mitigate risk effectively or face rejection in the payment market. 161 An Economic Perspective on the Enforcement of Credit Arrangements: The Case of Daylight Overdrafts in Fedwire Antoine Martin and David C. Mills
A fundamental concern for any lender is credit risk—the risk that a borrower will fail to fully repay a loan as expected. Thus, lenders want credit arrangements that are designed to compensate them for—and help them effectively manage—this type of risk. In certain situations, central banks engage in credit arrangements as lenders to banks, so they must manage their exposure to credit risk. This article discusses how the Federal Reserve manages its credit risk exposure associated with daylight overdrafts. The authors first present a simple economic framework for thinking about the causes of credit risk and the possible tools that lenders have to help them manage it. They then apply this framework to the Federal Reserve’s Payments System Risk policy, which specifies the use of a variety of tools to manage credit risk. The study also analyzes a possible increase in the use of collateral as a credit risk management tool, as presented in a recent proposal by the Federal Reserve concerning changes to the Payments System Risk policy.
Introduction
James McAndrews
James McAndrews
Introduction
imply put, payments settle things. A payment is typically theoretically and empirically; risk in retail payments systems; S the final step in a sequence of activities that make up an payments system development trends across countries; and the economic or financial trade. Because the parties to the trade interaction of the provision of reserves by central banks and the look forward to this final step, payments are the focus of their operation of payments systems. They illustrate the diversity of expectations about timing, risks, and costs. The more certain it interests and methods that economists have developed for is that this final step of a trade will take place, the easier it is for studying payment activities. trades to occur and for people to undertake efficient economic The contributions to this volume center on three broad actions. themes: theoretical models of money and payments, empirical As the world economy continues to grow, the share devoted analyses of trends in large-value payments, and risk to financial services has risen relative to the share representing management in payments systems. other economic activities. Payments have increased in The theoretical theme is examined in three articles—by importance as that trend has manifested itself, and their Morten L. Bech; Antoine Martin and James McAndrews; and contribution to the global economy is likely to increase as well. Todd Keister, Antoine Martin, and James McAndrews. The The rising importance of payments to economic activity in Bech study and the Martin-McAndrews study, both focusing general is a significant development—especially for banks and on large-value payments systems, explore the strategic central banks as providers of payment services. Bank customers incentives for banks to submit payments on time in different frequently rely on credit provided by their institutions to economic environments. They consider how the incentives are complete payments, whether they are using credit cards or affected by different central bank policies, such as the terms other payment instruments. Similarly, banks often rely on under which intraday credit is provided. Keister, Martin, and very-short-term credit provided by central banks to make McAndrews consider alternative models of monetary policy payments. Needless to say, whenever financial institutions implementation as well as the relationship between the provide credit, they must manage risk to prevent excessive risk demand for intraday balances to meet payment needs and the taking. Managing payments is therefore part of a larger risk reserve balances used to implement the policy objectives of the management process in the financial sector. monetary authorities. This special issue of the Economic Policy Review is devoted to Bringing an empirical perspective to the topic, Morten L. the economics of payments. The wide-ranging articles in this Bech, Christine Preisig, and Kimmo Soramäki conduct a global collection investigate large-value payments systems, both tour of developments in large-value payments over the last
James McAndrews is a senior vice president at the Federal Reserve Bank The views expressed are those of the author and do not necessarily reflect of New York. the position of the Federal Reserve Bank of New York or the Federal
FRBNY Economic Policy Review / September 2008 3 decade. Relying largely on data compiled by the Bank for Two studies consider the theme of risk management—one International Settlements, the authors discern a number of key by Michele Braun, James McAndrews, William Roberds, and trends in the growth and evolution of large-value payments. Richard Sullivan, the other by Antoine Martin and David C. In addition, two articles focusing on the timing of large-value Mills. Both apply the economic reasoning associated with risks payments systems—the first by Olivier Armantier, Jeffrey in credit arrangements to the specific case of payments. Braun Arnold, and James McAndrews and the second by Christopher et al. emphasize emerging retail systems while Martin and Mills Becher, Marco Galbiati, and Merxe Tudela—complement focus on the risk associated with a central bank’s intraday Bech’s theoretical work by explaining how different central lending to banks. bank policies influence payment timing. Both empirical The economics of payments is a rapidly developing field. analyses use data obtained directly from the large-value These studies offer a variety of approaches that use theoretical payments systems they study. Armantier, Arnold, and as well as empirical techniques to explore different aspects McAndrews review the timing distribution of payments of payments. Going forward, as researchers gain greater access in Fedwire, the Federal Reserve’s system, while Becher, to payment data, they will be better equipped to test other Galbiati, and Tudela consider the timing of payments in hypotheses about the determinants of behavior in payments CHAPS, the major U.K. system. The differences in timing systems. We hope that our findings will stimulate such between the two systems are found to be significant. initiatives and deepen interest in this dynamic field.
4 Introduction Theoretical Models of Money and Payments
Articles by:
Morten L. Bech Antoine Martin and James McAndrews Todd Keister, Antoine Martin, and James McAndrews
Morten L. Bech
Intraday Liquidity Management: A Tale of Games Banks Play
• To ensure smooth operation of real-time 1. Introduction gross settlement systems, central banks extend intraday credit, either against “[Banks] like to hang on to their cash and deliver it collateral or for a fee. as late as possible at the end of the working day.” “The Long Shadow of Herstatt,” The Economist, • As intraday credit is costly—either explicitly April 14, 2001 as fees or implicitly as the opportunity cost of collateral—participating banks seek to The value and volume of interbank payments increased minimize their use of liquidity by timing the dramatically throughout the 1980s and 1990s as a result of rapid release of payments. financial innovation and the integration and globalization of financial markets. In the United States, settlement of interbank payments grew from $300 trillion in 1985, or forty-five times • A game-theoretical study of intraday liquidity GDP, to almost $500 trillion in 1995, or seventy-five times GDP management behavior shows how the (Bech, Preisig, and Soramäki 2008). strategic incentives of banks depend on the Historically, interbank payments have been settled via intraday credit policy of central banks. deferred (end-of-day) netting systems. As the volume and value of transactions increased, however, central banks became • Two classic games emerge: “the prisoner’s worried about settlement risks inherent in netting systems. In dilemma” and “the stag hunt.” particular, the banks were concerned about the potential for contagion, or “knock-on,” effects attributable to the • The prisoner’s dilemma arises in a unwinding of net positions that would result if a participant collateralized credit scenario, where banks failed to meet its end-of-day obligations. Consequently, over delay payments even though they would be the last couple of decades, many countries have chosen to better off if they all sent payments early; the modify the settlement procedures employed by their interbank stag hunt arises in a priced credit scenario, payments system. where banks seek to coordinate the timing Most central banks opted for the implementation of a real- of their payments to avoid overdraft fees. time gross settlement (RTGS) system. By 1985, three central
Morten L. Bech is a senior economist at the Federal Reserve Bank of New York. The views expressed are those of the author and do not necessarily reflect the
FRBNY Economic Policy Review / September 2008 7 banks had implemented RTGS systems. A decade later, that different intraday credit policy regimes employed by central number had increased to sixteen, and at the end of 2006 the use banks. We characterize how the Nash equilibria depend on of RTGS systems had diffused to ninety-three central banks the underlying cost parameters, and discuss the efficiency (Bech and Hobijn 2007). implications of the different outcomes. As it turns out, two RTGS systems eliminate the settlement risk from unwinding classic paradigms in game theory emerge from the analysis: the because payments are settled irrevocably, and with finality, on “prisoner’s dilemma” and the “stag hunt.” Hence, many policy an individual gross basis and in real time. However, the questions can be understood in terms of well-known conflicts elimination of settlement risk comes at the cost of an increased and dilemmas in economics. This study uses the framework to need for liquidity to smooth nonsynchronized payment flows.1 conduct a comparative analysis of the relative desirability of Thus, central banks typically provide intraday credit. different intraday credit regimes from the perspective of a Two types of intraday credit policies have emerged among benevolent central bank. In addition, it discusses in turn how central banks: collateralized credit and priced credit (Furfine several extensions of the model will affect the results. These and Stehm 1998). Collateralized credit, in one form or another, extensions include settlement risk, incomplete information, is the prevailing option in Europe and elsewhere outside the heterogeneity, repeated play, multitudes of players, and more than just two actions. We conclude with general observations on the future of intraday liquidity management. This article develops a stylized game- theoretical model to analyze banks’ intraday liquidity management behavior 2. Intraday Liquidity Management in an RTGS [real-time gross settlement] Game
environment. Envision an economy with two identical banks using an RTGS system operated by the central bank to settle interbank claims.2 Bank A and Bank B seek to minimize the cost of making their United States. Collateralized credit usually takes the form of payments. We look at one business day that consists of two pledging collateral to the central bank or entering into an periods: morning and afternoon. intraday repurchase agreement with the central bank. Priced At dawn, both banks receive a request from a customer to credit is the policy of choice in the United States. The Federal pay $1 to a customer of the other bank on the same business Reserve has been charging a fee for intraday overdrafts since day.3 Assume for simplicity that the banks can either process 1994. Quantitative limits, or “caps,” are used in combination the request right away, or postpone it until the afternoon with both types of credit extensions. period. We abstract from reserve requirements and Intraday credit is costly, whether explicitly in the form of a fee precautionary motives for banks to hold balances with the or implicitly as the opportunity cost of the pledged collateral. central bank, and thus each bank has a zero balance on its Consequently, banks try to economize on their use of liquidity settlement account at dawn. Banks cannot send payments from throughout the day by carefully scheduling the settlement of their accounts in amounts that exceed their account balances. payment requests received from customers and the banks’ own However, banks can borrow funds from the central bank. The proprietary operations. Intraday liquidity management has cost of borrowing and how it is assessed depend on the intraday become an important competitive parameter in commercial credit policy of the central bank. Here, overdrafts are assessed banking and a policy concern of central banks (see, for example, at noon and at the end of the day. Overnight overdrafts are Greenspan [1996] and Berger, Hancock, and Marquardt [1996]). penalized at a very high rate, making banks avoid them This article develops a stylized game-theoretical model to altogether. analyze banks’ intraday liquidity management behavior in an If there were no adverse consequences, each bank would RTGS environment. It analyzes the strategic incentives under prefer to postpone making its payment and use the funds received via incoming payments from the other bank to 1 In most RTGS systems, payments are settled using reserve account balances at the central bank. For an individual bank, there are basically four different provide the balances to cover its own outgoing payments. sources of liquidity to fund outgoing payments: 1) overnight reserve balances, 2 In many countries, the interbank payments systems are neither owned nor 2) intraday credit extensions by the central bank, 3) borrowing from other operated by the central bank, but rather by a private company or a consortium banks via the interbank money market, and 4) incoming payments from other of banks. However, payments are usually settled in liabilities on the central banks. The first two sources affect the aggregate level of liquidity available in the bank. For ease of exposition, we ignore these differences here. system, while the latter two redistribute the liquidity among banks. Moreover, 3 liquidity from the first three sources generally comes at a price, whereas liquidity The customer could be internal to the bank, in which case the decision- from incoming payments is free from the perspective of the receiver. making agent can be thought of as the payment manager of the bank.
8 A Tale of Games Banks Play Game 1 Game 2 Intraday Liquidity Management Game Free Intraday Credit Game
Bank B Bank B morning afternoon morning afternoon morning 0, 0 0, D morning cA(m,m), cB(m,m) cA(m,a), cB(a,m) Bank A Bank A afternoon D, 0 D, D afternoon cA(a,m), cB(m,a) cA(a,a), cB(a,a) Note: We underline the cost associated with the strategy that is the best response of a bank given the strategy played by the other.
However, postponing is also costly, as customers might either demand compensation for late settlement or take their business outcomes by comparing both individual and aggregate elsewhere in the future, thereby imposing a reputation cost on settlement costs. The regimes are free, collateralized, or priced the delaying bank. intraday credit. For many payments, the cost of intraday delay is presumably small, as the underlying contractual obligation of the customer only specifies payment on a given business day. However, for an increasing number of financial transactions, the underlying contract stipulates payment prior to some 3. Free Intraday Credit Regime specific time on a given business day, and the cost of delay The first adopters of RTGS systems provided intraday credit could conceivably be high. Here, we simply assume the cost of for free, and we use this intraday credit policy regime as a delay to be a positive number D per dollar per period within.4 benchmark. With free credit within the day, there is no Moreover, postponing payments until the next day is extremely incentive to postpone payments. The free intraday credit game expensive in terms of reputation effects or direct compensation is shown in Game 2. to customers, so banks always submit any remaining payments It is best for both banks to play the morning strategy because in the afternoon. they incur no costs. Conversely, they incur the cost of delay if A convenient way of arranging the possible actions of the they postpone to the afternoon. The morning strategy banks and the associated costs is a 2 x 2 game, as shown in dominates the afternoon strategy, and the strategy profile Game 1. Each bank can play one of two strategies: morning or (morning, morning) is said to be an equilibrium in dominating afternoon. The first element in each cell denotes the settlement strategies. A pair of dominating strategies is a unique Nash cost of Bank A, whereas the second element denotes that of equilibrium. Bank B. Following Bech and Garratt (2003), we label this game In Game 2 and in other games, we adopt the convention of the intraday liquidity management game. underlining the cost associated with the strategy that is the best In the next three sections, we explore the games that emerge response of a bank given the strategy played by the other. We under different intraday credit regimes. Our solution concept summarize the results of the free intraday credit game in Result 1: is Nash equilibrium—that is, a set of strategies for which neither bank would wish to change its strategy on the Result 1: Early settlement (morning, morning) is a unique assumption that the other bank will not change its strategy equilibrium in the free intraday credit game. The outcome is either. We focus on Nash equilibria in pure strategies—that is, efficient in that it ensures the lowest possible aggregate strategies where a player chooses to take one action with settlement cost across all pairs of strategies. probability 1—which is in contrast to a mixed strategy, where individual players choose a probability distribution over In reality, payment flows are not perfectly symmetric as they several actions. We evaluate the efficiency of different are in the model, and imbalances frequently occur. Moreover, the zero price for intraday credit creates no incentives to 4 Delay has several private and social costs associated with it. First, time is economize on overdrafts.5 In fact, the size of the overdrafts money (even intraday) and hence delay of settlement may displease customers and counterparties, which are left with higher costs and greater uncertainty. generated by banks (relative to their capital base) in an RTGS Second, delayed settlement increases operational risk insofar as the time span environment came as a surprise to many central banks. As during which an incident may disrupt the settlement process increases and the guarantor of the finality of payments, the central bank is time to recover after an incident decreases. Third, the process of delaying can be costly, and the resources devoted to managing intraday positions are a cost. exposed to credit risk—as, ultimately, are taxpayers. Hence, Fourth, delay increases the length of time participants may be faced with credit central banks are almost unanimous in the opinion that the risk exposures vis-à-vis each other. provision of free intraday liquidity is not a viable option.6
FRBNY Economic Policy Review / September 2008 9 4. Collateralized Intraday Credit Game 3 Regime Collateralized Credit Game Bank B In most countries, central banks provide commercial banks morning afternoon with intraday credit against collateral. The practical morning C, C 2C, D Bank A implementation varies across countries, depending on the afternoon D, 2C C + D, C + D institutional infrastructure for the safekeeping and settlement of securities. For ease of exposition, we assume that credit is extended via intraday repurchase agreements (repos), as in the United Kingdom and Switzerland. the cost of delaying is greater than the cost of obtaining Under the intraday repo agreement, the central bank liquidity—that is, D > C—then banks have no incentive to provides the bank with $1 in its account at the beginning of the delay and the strategy profile (morning, morning) is the only period in return for eligible collateral worth the same amount Nash equilibrium. If Bank B plays morning, the best strategy plus a “haircut” to cover any market and credit risk associated for Bank A is to play morning as well. Moreover, if Bank B with the collateral. At the end of the period, the transaction chooses to postpone, the best strategy for Bank A is still is reversed. The central bank does not charge explicit interest morning. In other words, morning is a dominating strategy for this service, but the collateral subject to repo entails an for Bank A and, by symmetry, for Bank B as well. However, opportunity cost for the banks, as this collateral cannot be used if the cost of liquidity is higher than the cost of delaying— for other purposes. The opportunity cost of collateral is that is, C > D—then the strategy profile (afternoon, assumed to be C per period per dollar. afternoon) is the only Nash equilibrium. It is a unique Nash If Bank A and Bank B both decide to process their requests equilibrium, since neither bank wishes to switch to morning early, then they each have to engage in an intraday repo with if the other bank keeps playing afternoon because a switch the central bank in order to obtain liquidity, and consequently would increase its settlement cost. However, it is also clear they will each incur the cost C. However, if, say, Bank A decides that the banks would be better off if they both chose to to delay while Bank B decides to process, then Bank A will incur process payments in the morning. Unfortunately, (morning, the cost of delay D in the morning period. However, in the morning) is not an equilibrium in this one-shot game. afternoon period, it can use the incoming liquidity from Bank Starting from (morning, morning), each bank would wish to B to fund its own outgoing payment in the next period. postpone payment in order to lower its settlement cost. This Conversely, Bank B receives no liquidity and has to roll over the strategic situation is a classic paradigm in game theory called repo with the central bank for an additional period and incur the prisoner’s dilemma.7 We summarize the results of the the cost C one more time for a total of 2C. Finally, if both banks collateralized credit game in Result 2: choose to postpone, they both incur the cost of delay D. Result 2: In the collateralized credit game, early settlement Moreover, at noon they still have no liquidity available, and (morning, morning) is a unique equilibrium if the both have to engage in an intraday repo in the afternoon period opportunity cost of collateral is less than the cost of delaying for which they each will incur the opportunity cost of collateral C. The settlement costs are summarized in Game 3, hereafter 7 The “prisoner’s dilemma” is the most famous paradigm in game theory. referred to as the collateralized credit game. Suppose that the police have arrested two former felons who they know have In the collateralized credit game, the equilibrium committed an armed robbery together. Unfortunately, they lack enough admissible evidence to get a jury to convict them of armed robbery. They do, depends solely on the relative size of the opportunity cost of however, have enough evidence to send each prisoner away for two years for collateral and the cost of postponing a payment request. If theft of the getaway car. Prisoner’s Dilemma 5 In 2006, the Federal Reserve eliminated the extension of free intraday credit to government-sponsored enterprises (GSEs) and certain international Prisoner 2 organizations for the purpose of securities-related interest and redemption Confess Silence Confess 5, 5 0, 10 payments. This action was taken in part because, for some issuers, the lag Prisoner 1 between the time the Federal Reserve credited the interest and redemption Silence 10, 0 2, 2 payments to the recipients’ accounts (early in the morning) and the time the issuer covered the resulting overdraft extended, at times, until shortly before The chief inspector now makes the following offer to each prisoner: “If you will the close of the Fedwire system, hence exposing the Federal Reserve to credit confess to the robbery, implicating your partner, and he does not also confess, risk for the duration of the day. Currently, interest and redemption payments then you’ll go free and he will get ten years. If you both confess, you’ll each get have to be funded up front. five years. If neither of you confesses, then you’ll each get two years for the auto 6 In models without credit risk, Freeman (1996) and Martin (2004) find that theft.” It is a Nash equilibrium for each prisoner to confess; yet they would both free intraday credit is the socially optimal policy. be better off if they both chose to remain silent.
10 A Tale of Games Banks Play Chart 1 Game 4 Pricing Structure for Swiss Interbank Clearing Offsetting in the Morning Game (C > D) CHF per payment 3.5 Bank B morning afternoon 3.0 Sender (above morning 0, 0 2C, D 100,000 CHF) Bank A afternoon D, 2CC + D, C + D 2.5 Note: We underline the cost associated with the strategy that is the best 2.0 response of a bank given the strategy played by the other. Sender (under 1.5 100,000 CHF) 1.0
0.5 Receiver Third, central banks can use pricing. For example, the Swiss 0 Before 8 a.m. 8 a.m. - 11 a.m. 11 a.m. - 2 p.m. After 2 p.m. National Bank charges higher prices for payments sent later in the day, thereby giving banks a direct incentive to process early. Source:
8 See Danmarks Nationalbank (2003).
FRBNY Economic Policy Review / September 2008 11
Game 5 Game 6 Priced Credit Game Priced Credit as the Stag Hunt Game
Bank B Bank B morning afternoon morning afternoon morning 0, 0 F, D morning 0, 0 5, 2 Bank A Bank A afternoon D, F D, D afternoon 2, 5 2, 2
Note: We underline the cost associated with the strategy that is the best response of a bank given the strategy played by the other.
dominates late submission, and one would expect banks to choose the cost-efficient strategy. In sum, the introduction of if both banks choose to postpone, neither bank would a gridlock-resolution mechanism may change submission unilaterally want to deviate and process because that would behavior. We offer the following conjecture: increase its settlement cost from 2 cents to 5 cents. Conjecture: Gridlock resolution and offsetting mechanisms Here, the priced credit game has the structure of a classic 9 may eliminate the potential prisoner’s dilemma. coordination game called the stag hunt. The key feature of the stag hunt game is that while the (morning, morning) The issue of liquidity-saving mechanisms is discussed equilibrium is preferred by both players in terms of lowest cost, further in Martin and McAndrews (2008). the other is preferred in terms of strategic risk. In the early- settlement equilibrium, the settlement cost of one bank depends on the action of the other. One bank’s deviation from morning, for whatever reason, will impose increased 5. Priced Intraday Credit Regime settlement costs on the other bank. In contrast, the strategy to postpone payments carries no risk in the sense that the Under the priced credit regime, banks are charged the fee F per settlement cost is the same regardless of which action the other dollar if their settlement account is overdrawn at the end of a bank takes. A cautious bank may reasonably choose to period. This implies that no overdraft fee is incurred if the postpone, ensuring the 2 cents with certainty rather than banks manage to synchronize their payments. The settlement risking the cost of 5 cents. This is especially true if there are costs associated with the different possible pairs of strategies concerns regarding the other bank’s ability to coordinate (for are shown in Game 5, the priced credit game. example, because of operational risk). We recap the results of If both banks play morning, then payments net out and the priced credit game in Result 3: banks incur no costs. The payments also net out if both banks play afternoon, but each will incur the cost of delay. If one bank Result 3: In the priced credit game, early settlement pays and the other delays, then the paying bank will incur (morning, morning) is a unique equilibrium if the an overdraft at noon while the other can use the incoming overdraft fee is less than the cost of delaying (F < D). The payment from the morning period to fund its outgoing outcome is efficient. In contrast, both (morning, morning) payment in the afternoon. However, the bank that delays will and (afternoon, afternoon) are feasible equilibria if F > D incur the cost D. and the game is a stag hunt. Late settlement is inefficient. As in the collateralized credit regime, the outcome depends In the prisoner’s dilemma, there is a conflict between on the relative size of the cost of liquidity and the cost of individual rationality and mutual benefit. In the stag hunt, postponing the processing of a request. Again, the strategy rational players are pulled in one direction by consideration of profile (morning, morning) is a unique Nash equilibrium if the mutual benefit and in the other by individual risk concerns cost of liquidity is less than the cost of delaying the payment (Skyrms 2004). In the stag hunt game, the outcome depends on request—that is, F < D. the player’s appetite for strategic risk—that is, the uncertainty However, if F > D, then the strategy profiles (morning, morning) and (afternoon, afternoon) are both Nash equilibria. To 9 The “stag hunt” is a story that became a game. The game is a prototype of the see this, assume that F = 5 cents and D = 2 cents, as in Game 6. social contract. The story is briefly told by the eighteenth-century philosopher If both banks choose to process payments early, then neither Jean-Jacques Rousseau in A Discourse on Inequality (Skyrms 2004): “If it was a matter of hunting a deer, everyone well realized that he must remain faithful to bank would want to change and postpone because that would his post; but if a hare happened to pass within reach of one of them, we cannot increase its settlement cost from 0 cents to 5 cents. Likewise, doubt that he would have gone off in pursuit of it without scruple.”
12 A Tale of Games Banks Play that arises from the interaction between the players of the Chart 2 game. The conflict in the priced credit game is a trade-off Comparative Analysis of Intraday Credit Regimes between lower settlement costs and strategic risk. Priced credit One way of pinning down a unique equilibrium is by using Harsanyi and Selten’s (1988) concept of risk dominance.10 In a F F = 2C F = C symmetric 2 x 2 game, risk dominance asserts that players will 4: 3: choose the strategy that gives the highest expected payoff under Collateralized Priced the assumption that the opponent randomizes with equal (a,a) credit credit preferred preferred probability over the two available strategies. Fixing D to be 2 cents in Game 6 implies that (morning, morning) is the risk- 2D dominant equilibrium if F < 4 cents and (afternoon, afternoon) 1: is the outcome if F > 4 cents. Priced credit preferred Result 4: In the priced credit game, the risk-dominant 2: Priced equilibrium is early settlement (morning, morning) if (m,m) credit F < 2D. Otherwise, late settlement (afternoon, afternoon) is preferred the risk-dominant equilibrium.
Using the analysis above, we now turn to a comparison of the aggregate settlement costs to the economy under 45° collateralized and priced intraday credit policy regimes. (m,m) D (a,a) C Collateralized credit
6. Choice of Intraday Credit Policy
An omnipresent question for central banks is the choice of The aggregate settlement costs under the equilibria of the intraday credit policy. For example, the Board of Governors of the two intraday credit regimes are easily calculated by summing Federal Reserve System is currently reviewing its Payment System the entries in each cell in Game 3 and Game 5, respectively. The Risk Policy with a view to reducing liquidity risk, credit risk, and aggregate settlement costs when one bank is playing morning operational risk while maintaining or improving payments system and the other afternoon are 2C + D and F + D, respectively, efficiency. For this reason, the Federal Reserve Board published under the two regimes. With two possible (risk-dominant) a consultation paper in June 2006 to elicit information from Nash equilibria under each regime, there are four different financial institutions and other interested parties on their scenarios to consider. experiences managing intraday risk associated with Fedwire funds The comparative analysis is summarized in Chart 2. The X-axis transfers. Our model can provide insight into the desirability of shows the Nash equilibrium in a collateralized credit regime different payments system policies and highlight some of the as a function of the opportunity cost of collateral. The Y-axis difficulties facing policymakers. shows the risk-dominant Nash equilibrium under a priced Assume that a central bank is a benevolent provider of the credit regime as a function of the overdraft fee. In the priced RTGS system insofar as it seeks to secure the lowest possible credit regime, aggregate settlement costs are zero if the aggregate settlement costs for the economy. The central bank equilibrium is (morning, morning). In contrast, a collateralized can choose between a collateralized credit and a priced credit credit regime always implies positive settlement costs. regime. Additionally, for the purposes of this analysis, assume Consequently, priced credit is the preferred regime if F < 2D— that the central bank cannot (further) influence the cost of that is, in scenarios 1 and 2. In other words, take the parameters liquidity under either regime or the cost of delay. The preferred of the model as exogenously given. regime then depends on the equilibrium outcome under the If payments are delayed under both regimes—that is, 2D two regimes, which in turn depends on the relative magnitudes < F and D < C—then aggregate settlement costs are 2D + 2C of F, C, and D. and 2D, respectively. Hence, priced credit is the preferred regime C in scenario 3 in Chart 2. Conversely, collateralized 10 In 1994, John C. Harsanyi and Reinhard Selten were awarded the Nobel Prize in Economics, together with John F. Nash Jr., for their pioneering analysis of credit is the preferred regime if banks do not delay payments equilibria in the theory of noncooperative games. under such a regime but they do under a priced credit
FRBNY Economic Policy Review / September 2008 13 regime—that is, C < D < 2D < F (scenario 4). We summarize 7. Settlement Risk this as: Settlement risk is an important concern in all payment Result 5: Priced credit is preferred to collateralized credit arrangements (see, for example, Kahn, McAndrews, and except when collateralized credit leads to quicker settlement Roberds [2003] and Mills and Nesmith [2008]). Fundamentally, of payments. it is the risk that settlement does not take place as expected. The model provides very clear results in terms of the As such, settlement risk comprises both liquidity and credit desirability of the two regimes, but in reality the analysis is risks. Liquidity risk is the risk that a counterparty will not settle more involved. Moreover, the analysis does not take into an obligation for full value when due, but at some unspecified account default risk, against which the collateral protects. time thereafter. Credit risk is the risk that a counterparty A challenge for comparative analysis in practice is that the cost will not settle an obligation for full value either when due or of delay is not observable. In fact, little is known about the anytime thereafter. The presence of settlement risk affects the costs banks face if they delay settlement of payments. Without strategic interaction between banks and hence their intraday knowledge of the cost of delay, the comparative analysis liquidity management behavior. becomes less informative, but the simple analysis presented in To illustrate how settlement risk affects strategic Chart 2 does yield the following necessary, but not a sufficient, interaction, we assume that the banks have entered into condition for collateralized credit to be the preferred regime: trades with each other yielding obligations to pay the other $1, to be settled gross. On settlement day, a bank might Result 6: For collateralized credit to be the preferred regime, a necessary condition is that the opportunity cost of collateral be lower than (literally half) the overdraft fee The presence of settlement risk affects the charged under priced credit. strategic interaction between banks and The opportunity cost of collateral is not directly observable hence their intraday liquidity management either, but the rate differential between federal funds loans, which are uncollateralized, and loans through repurchase agreements, behavior. which are collateralized, suggests that the opportunity cost is in the range of 12 to 15 basis points per annum (see Board of Governors experience an operational incident or default altogether, of the Federal Reserve System [2006]).11 However, the overdraft which leads to either a temporary inability or permanent fee is readily observable because it is set by the central bank with failure to pay. Because payment flows are symmetric, a view to managing credit exposure from overdrafts. Currently, neither bank starts out with an exposure vis-à-vis the other daylight overdraft fees in the Fedwire Funds Service are calculated at dawn. If a bank defaults before the opening of the RTGS using an annual rate of 36 basis points, quoted on the basis of a system, the other bank can just withhold its payment. 21.5-hour day. This simple “back-of-the-envelope” comparison However, by paying early, a bank exposes itself to the suggests that there may be scope for investigating an increased inability or failure of the other to pay. Everything else being role for collateral in the Fedwire system. equal, one would expect banks to be more cautious in their In the following sections, we investigate how the conclusions behavior when facing settlement risk. In essence, settlement from the model are likely to change as more realism is added. We risk reduces the effective cost of delaying. start by considering settlement risk, followed by incomplete Specifically, we model liquidity risk by assuming that, with information, repeated play, and more than two banks and periods. probability ω, banks will not be able to submit payments to the In Box 1 and Box 2, we analyze, respectively, the strategic RTGS system in the morning because of, say, a telecommuni- interaction between banks when there is no intraday credit cations outage. However, the telecommunications links to available and when banks are heterogeneous. stricken banks are reestablished at noon and banks can then make payments in the afternoon period. The expected costs for banks are derived in the appendix. These costs are used in the intraday liquidity management games with liquidity risk shown in Game 7 and Game 8 for the two policy regimes (collateralized credit and priced credit). For convenience, we 11 The opportunity cost of collateral would, in all likelihood, increase if the Federal Reserve implemented a collateralized credit regime because the show only the costs for Bank A, but by symmetry the costs are demand for collateral would increase. the same for Bank B.
14 A Tale of Games Banks Play Box 1 The No Intraday Credit Game
The exception proving the rule that early adopters provided than the cost of delay, the game is an anti-coordination game, so intraday credit for free is Switzerland. The Swiss National Bank named because it is mutually beneficial for the players to play implemented real-time gross settlement (RTGS) systems in 1987, different strategies. If Bank B plays morning, then the best response but did not provide intraday credit until the autumn of 1999. The of Bank A is to play afternoon. Conversely, if Bank B plays change in policy was motivated by an increase in time-critical afternoon, then the best strategy for Bank A is to play morning. The payments and, in particular, the future introduction of the underlying conflict in the game is that both banks want to benefit Continuous Linked Settlement system for foreign exchange from free liquidity, but liquidity is rivalrous—that is, banks cannot transactions. According to Heller, Nellen, and Sturm (2000), benefit from it at the same time. Hence, both (morning, afternoon) the amount of payments settled by noon rose from one-third and (afternoon, morning) are possible Nash equilibria, but neither to one-half of the daily turnover as a result. Pareto dominates the other or is focal in any sense. The mixed- Going against conventional wisdom, the Reserve Bank of strategy Nash equilibrium implies that banks play morning with New Zealand implemented a new liquidity management regime in probability pD= ⁄ ρ and afternoon with the complementary 2006 that discontinued its intraday automatic reverse repurchase probability. The expected settlement cost for a bank is ρ (see facility (autorepo). Instead, the Reserve Bank chose to supply a appendix). Hence, the mixed strategy does not Pareto dominate significantly higher level of cash (overnight monies) sufficient to either of the pure Nash equilibria, and a bank might as well play enable participants to settle payments efficiently. The change was morning and save itself the trouble of randomizing. necessitated by a growing scarcity of New Zealand government It is not obvious how banks can solve the conundrum of who securities (see Reserve Bank of New Zealand [2006]). gets the benefit of free liquidity. One solution in these types of If the central bank does not provide intraday credit, then games is for banks to engage in pre-play communication. In pre- payments have to be funded by balances held with the central bank, play communication, each player announces the action it intends interbank money market borrowings, or incoming payments from to take (or, alternatively, the action it would like the other to take). other banks. The first two sources are costly, whereas the last is free In game theory, pre-play communication that carries no cost is from the perspective of the receiver. Let ρ denote the (marginal) referred to as cheap talk (Farrell and Rabin 1996). Interestingly, opportunity cost of balances held at the central bank. The in some experimental settings, cheap talk has been found to be opportunity cost of reserves is closely linked to the central bank’s effective. Another form of pre-play communication is for one bank policy with respect to remunerating reserves. If the central bank to signal convincingly that it will play afternoon. One way to do does not pay interest on reserves, then the opportunity cost is close this would be to open late, but that would probably be bad for to the overnight money market rate, whereas if the central bank business in general and thus costly. does pay interest on reserves, it depends on the difference between Aumann (1974) provides a generalization of Nash equilibrium the money market rate and the administrative rate paid on known as correlated equilibrium, which allows for possible reserves. dependencies in strategic choices. A perfectly correlated The no intraday credit game is given below for the interesting equilibrium would be for banks to use a fair coin to determine case where ρ > D . which bank gets to play afternoon. In a repeated setting, a convention for banks to alternate sending early could conceivably No Intraday Credit Game ( > D) evolve. Above and beyond the potential instability of the equilibrium Bank B outcome, a key insight of the no intraday liquidity management morning afternoon game is that the monetary policy stance may directly affect the morning , , D Bank A afternoon D, D + , D + settlement of payments intraday owing to the close link between the opportunity cost of holding reserves and the overnight interest Note: We underline the cost associated with the strategy that is the rate. Any movement in the monetary policy stance will affect the best response of a bank given the strategy played by the other. opportunity cost and may shift the equilibrium around. Interestingly, Heller, Nellen, and Sturm (2000) claim that a less If the opportunity cost of reserve balances is less than the cost restrictive monetary policy stance from 1993 to 1999 can explain a of delay, then (morning, morning) is the equilibrium in dominating large part of the reduced congestion observed in Swiss Interbank strategies. However, if the opportunity cost of reserves is larger Clearing, as this led banks to hold increased account balances.
FRBNY Economic Policy Review / September 2008 15 Box 2 Heterogeneity
In the analysis, we focus on the interaction between two identical loan late is a weakly dominating strategy because if the issuer for banks. Obviously, participants in a real-time gross settlement some reason should delay then it would be best for the bank to system are not a homogenous group. Here, we explore the delay as well. A small intraday opportunity cost for a bank using implications of introducing heterogeneity among participants. For overdraft capacity to cover the interest on the loan would eliminate ease of exposition, we consider only two cases. First, we look at the the (morning, morning) equilibrium. The (morning, afternoon) case where participants face different liquidity and delay costs. We equilibrium leaves the issuer with an overdraft at the central bank do this in the context of a recent policy change in the Fedwire for the entire day. In sum, different cost structures for participants system. Second, we consider the case where payment flows are not can lead to interesting games with asymmetric equilibria. We now balanced and then we try to gauge the extent to which that affects turn to payment flow imbalances. the strategic interaction between banks. On any given day, payment flows are never balanced because In 2006, the Federal Reserve eliminated the extension of free banks receive different amounts of payment requests from their intraday credit to government-sponsored enterprises (GSEs) and customers. Banks manage their projected end-of-day balances certain international organizations for the purpose of securities- throughout the day. Liquidity is redistributed via the interbank related principal and interest payments. This action was taken in part money market from the “haves” to the “have nots.” The question because, for some issuers, the lag between the time the Federal is the extent to which such differences in payment flows can affect Reserve credited the interest and redemption payments to the the strategic interaction among banks. To provide insight, we recipients’ accounts (early in the morning) and the time the issuer assume that Bank B has two $1 payments to send to Bank A covered the resulting overdraft extended, at times, until shortly whereas Bank A still has only $1 to send to Bank B. The strategy set before the close of the Fedwire system. As a result, the Federal expands for Bank B, which can choose to send them both early, Reserve was exposed to credit risk for the duration of the day. delay them both, or send one early while holding back the other. Currently, principal and interest payments have to be funded up The resulting games are shown below. front. To see how the simple framework can account for this Collateralized Credit Game with Payment Flow observation, assume that one player is now an issuer of securities Imbalance (C > D ) and needs to pay $1 in principal and interest to the other player— a bank. Assume that issuer had the necessary cash to pay out Bank B principal and interest on hand the previous day and chose to lend m,m m,a a,a mC, 3CC, D+2C 2C, 2D+C it out in the overnight money market to earn a return. For Bank A simplicity, also assume that the borrower was the bank that aD, 4CD, D+3CD+C,2D+2C henceforth has to return the $1 plus interest ()ρ to the issuer. Note: We underline the cost associated with the strategy that is the The central bank is granting free intraday credit to the issuer best response of a bank given the strategy played by the other. but charges the bank for overdrafts. Owing to market conventions, the cost of delaying the payout of principal and interest is high (H), The games become slightly more complicated, but the whereas the cost of delaying the return of a money market loan to fundamental issues remain. In the case of collateralized credit, a participant that has access to free intraday credit is virtually nil. banks may still end up delaying payments even though it is more The resulting principal and interest game is shown below. efficient to process early in terms of minimizing aggregate settlement cost. In the case of priced credit, it is possible only to offset two payments against each other, and thus it turns out that Principal and Interest Game Bank B will always hold back one payment. The stag hunt is played with the remaining payment. Issuer morning afternoon Priced Credit Game with Payment Flow morning F, 0 (1+ )F, H Bank afternoon F, 0 F, H Imbalance ( F > D )
Note: We underline the cost associated with the strategy that is the Bank B best response of a bank given the strategy played by the other. m,m m,a a,a m 0, 2F 0, D+F F, F+2D Bank A Clearly, it is a dominating strategy for the issuer to pay out aD, 3FD, D+2FD, 2D+F early. If the issuer plays morning, then the bank is indifferent Note: We underline the cost associated with the strategy that is the between returning early or late. However, returning the overnight best response of a bank given the strategy played by the other.
16 A Tale of Games Banks Play Game 7 Exhibit 2 Collateralized Credit Game with Liquidity Risk Equilibrium and Efficiency in the Priced Credit Game with Liquidity Risk Bank B morning afternoon (a, a)(m,m), (a, a)(m,m) morning (1+ )C+ D 2C+ D Bank A afternoon D+ CD+C efficient stag hunt efficient
F F D Game 8 Priced Credit Game with Liquidity Risk
Bank B Game 9 morning afternoon morning (1- )F+ D(1- )F + D Collateralized Credit Game with Credit Risk Bank A afternoon D D Bank B morning afternoon morning C (2- )C+ (1- ) Bank A afternoon (1- )D (1- )(C+D) Exhibit 1 Equilibrium and Efficiency in the Collateralized Credit Game with Liquidity Risk Game 10 Priced Credit Game with Credit Risk
(a, a)(a, a)(m,m) Bank B efficient prisoner’s dilemma efficient morning afternoon morning 0 (1- )F+ (1- ) Bank A afternoon (1- )D (1- )D C C D 1- 1-
The introduction of liquidity risk implies that a bank risks bank will not be making any further payments. Thus, the other incurring the cost of delay even if it is playing the morning strategy. bank has to borrow from the discount window at rate R in On the flip side, the other bank incurs additional liquidity costs order to square its account at the end of the day. Furthermore, due to the lack of an incoming payment. As such, liquidity risk assume that a surviving bank eventually will recover ()1 – α of affects both the equilibrium outcomes and the efficiency thereof. the dollars that it is owed. Hence, the total cost of default is In a collateralized credit regime, the (afternoon, afternoon) θα= + R . equilibrium becomes more likely (Exhibit 1). We assume that if a bank defaults then there is no reputation Without liquidity risk, the condition for late settlement is cost of delaying. Hence, the expected cost of delay is ()1 –δ D. D < C, whereas with liquidity risk it is D < C/(1-ω ). Increasing The resulting games for the collateralized and priced credit the exposure to liquidity risk—that is, ω→1 — raises the regimes are shown in Game 9 and Game 10, respectively. likelihood that banks hold back payments as C ⁄ ()1 –ω → ∞ . Again, the settlement costs are derived in the appendix, and Interestingly, holding back payments is the efficient outcome if only the expected settlement costs for Bank A are shown. the cost of delay is sufficiently low or the risk of a temporary For collateralized credit, it turns out that the results are failure to pay is high. The prisoner’s dilemma disappears if identical to those for liquidity risk. The only difference is that the probability of a default, δ , replaces the probability of a D <ω C ⁄ ()1 –ω . temporary failure to pay, ω , in Exhibit 1. With priced credit, With priced credit, banks will still play (morning, morning) banks will play the (morning, morning) equilibrium whenever if the cost of liquidity is less than the cost of delay; and if D DF> + δθ compared with D > F, when there is no risk of drops below F, the stag hunt emerges, as in Result 3 (Exhibit 2). default, as in Result 3. Otherwise, the game is a stag hunt. We However, (afternoon, afternoon) is now a feasible unique Nash sum up the results from introducing settlement risk as: equilibrium. This equilibrium is efficient and will be the outcome if the exposure to liquidity risk is sufficiently high— Result 7: Settlement risk makes (other things being equal) that is, ω > DF⁄ . late settlement (afternoon, afternoon) a more likely To model credit risk, we assume that a bank with probability outcome of the intraday liquidity management game. δ will be closed by its regulator at noon. Should that occur, the Late settlement may be efficient.
FRBNY Economic Policy Review / September 2008 17 8. Incomplete Information Game 12 Priced Credit Game with Incomplete Information The analysis so far has assumed that banks have complete Bank B information with regard to the payments to be settled. In morning afternoon reality, banks have only an incomplete picture during the day. morning (1-p)F F Bank A In fact, there can be substantial uncertainty about both afternoon D D incoming payments and requests that customers will submit over the remainder of the day. This ambiguity further complicates the task of managing the liquidity position of a Exhibit 3 bank within the day. Equilibrium and Efficiency in the Bayesian Bech and Garratt (2003) develop a Bayesian game in which Collateralized Credit Game banks have private knowledge about their own pending payment requests but only imperfect information about those (a,a)(a,a)(m,m) of the opponent. Moreover, banks face uncertainty inefficient prisoner’s dilemma efficient (fundamental) about the arrival of new payment requests and
uncertainty (strategic) in terms of the opponent’s action. In the (1-p) C C D model, payment requests arrive from customers at dawn and at noon with probabilities p and q, respectively. Banks seek to minimize expected settlement costs. It turns out that the strategies of banks are determined by the action they take when In the case of collateralized credit, it is still true that they do receive a request at dawn. This simplifies the analysis (morning, morning) is the Nash equilibrium whenever D > C and allows us to stay within a 2 x 2 framework for the purposes because the additional cost of (1-p)C is incurred regardless of of determining equilibria. We construct games where the whether the bank in question is playing morning or afternoon. payoffs are conditional on having received a request from the Otherwise (afternoon, afternoon) is the equilibrium (Exhibit 3). perspective of each bank. For example, the expected settlement While Game 11 is only a prisoner’s dilemma when the costs of sending early against an opponent that also sends early additional cost of delaying is larger than the expected cost of (if it has a request) are C + (1–p)C and (1–p)F, respectively. The processing the payment—that is, (1–p)C < D < C —it is still extra component relative to collateralized and priced credit inefficient, from an aggregate expected settlement cost games described earlier reflects the chance that the opponent perspective, to delay. Early settlement is the only efficient might not have received a request and hence the bank would outcome. have to borrow additional liquidity from the central bank. The In the case of priced credit, (morning, morning) is again the outcomes under the two intraday credit regimes—now Bayes- equilibrium whenever D > F. Conversely, if D < (1 - p)F, then Nash equilibria—are determined by Game 11 and Game 12, the strategy profile (afternoon, afternoon) is the Bayes-Nash respectively. However, the full Bayesian game is needed to equilibrium (Exhibit 4). In the intermediate case, the stag hunt evaluate the efficiency implications of different strategy profiles emerges once again. In contrast to the case of collateralized as banks individually do not take into account the positive credit, it is possible for postponement of payments to be externality of liquidity to the other bank. We cite the results on efficient in the case of priced credit. This occurs if D < (1-p)qF efficiency here and refer the reader to the original paper (Bech —that is, if the cost of delay is low (relative to the overdraft fee) and Garratt 2003) for the details. and the arrival of payment requests is sufficiently skewed toward the afternoon (low p and high q). In that case, the expected benefit from being able to offset payments in the afternoon outweighs the cost of delay. We reiterate the outcome of introducing incomplete Game 11 Collateralized Credit Game with Incomplete information and random arrivals of payment requests in the Information following result:
Bank B Result 8: Incomplete information about payment flows morning afternoon increases (other things being equal) the likelihood of late morning (2-p)C 2C Bank A settlement (afternoon, afternoon). In the case of priced afternoon D + (1-p)C C + D credit, late-day requests may make late settlement efficient.
18 A Tale of Games Banks Play Exhibit 4 Game 13 Equilibrium and Efficiency in the Bayesian Repeated Collateralized Credit Game Collateralized Credit Game Bank B (a,a)((a,a) m,m), (a,a)(m,m) trigger always delay C β()CD+ efficient inefficient stag hunt efficient trigger ------2C + ------1 – β 1 – β Bank A (1-p)qF (1-p)FFD β()CD+ CD+ always delay D + ------1 – β 1 – β
9. Repeated Interaction Game 14 In most payments systems, participating banks interact Repeated Priced Credit Game repeatedly with each other, both within and across days. It is Bank B well known that playing the same game (such as the prisoner’s trigger always delay dilemma) numerous times might yield a different outcome. trigger 0 F + D/(1- ) Bank A Unlike a game played once, a repeated game allows for a always delay D/(1- )D/(1- ) strategy to be contingent on past moves, thus allowing for reputation effects and retribution. The key is that cooperation now can be rewarded by cooperation later, and cheating can be punished by not cooperating later. It is thus not always wise where t denotes the trigger strategy and bank iAB∈ {}, . In the to pursue a short-run gain in a repeated game. case of priced credit, the future discounted settlement costs for A trigger strategy in which cheating is punished in a bank playing trigger strategy against an opponent that always subsequent periods can encourage cooperation. One harsh delays are the overdraft fee in the first period and then the cost example is for a player to begin by cooperating in the first of delay for any subsequent days in which they interact. That is: period and to continue cooperating until a single defection by β i(), β β 2 β 3 …= F + ------D, the opponent, after which the player never cooperates again. c ta = F ++++D D D 1 – β A less harsh trigger strategy is “tit for tat,” where a player where a denotes the “always delay” strategy. The settlement responds in one period with the same action the opponent used cost for the remaining strategy profiles can be derived in a in the last period. The repetition of a game may solve some of similar fashion. The resulting games are shown in Game 13 and the single-play issues discussed above. However, by offering Game 14, respectively. We show only the discounted future more complex strategies, a repeated game can also result in settlement cost for Bank A. The prisoner’s dilemma remains in more equilibrium outcomes. In other words, the repetition of the collateralized credit regime if the future matters little to a game itself does not necessarily solve the quandaries faced banks, as illustrated in Exhibit 5. In fact, with β = 0, we get the by players in single-play games. Additional structure is often one-stage collateralized credit game described earlier. However, if needed. the discount factor is significantly large—that is, β > 1 – DC⁄ — Here, we assume an infinite play setting where banks then repeated play transforms the prisoner’s dilemma into a stag discount the future. The daily discount factor is given by hunt. Moreover, as shown in the appendix, if the discount <<β 0 1 . Banks can choose between two possible strategies. factor is even higher—that is, β > 2()CD– ⁄ ()2CD– —then One strategy is to always delay. The other is a trigger strategy the risk-dominant equilibrium is (trigger, trigger). whereby a bank will send early as long as the other does, but will In the case of priced credit, the infinitely repeated version delay afterward if the other bank deviates. Using the formula of the game remains a stag hunt game if F > D. However, for infinite geometric series, we can compute the future early processing is the risk-dominant equilibrium if discounted settlement cost under the two strategies for each of F < ()2 – β D ⁄ ()1 – β compared with F < 2D in the one-stage the two intraday credit regimes. For example, in a collateralized game (see appendix). Hence, the more the future matters for credit regime where both banks are playing trigger, the future banks, the more likely it becomes that banks will coordinate discounted settlement costs are: toward early processing. We summarize the results of i 2 3 C c ()tt, = C ++βC β C +β C +… = ------, 1 – β introducing repeated play in Result 9:
FRBNY Economic Policy Review / September 2008 19 Exhibit 5 Game 15 Equilibria and Efficiency in the Repeated Three-Player Game with Collateralized Credit Collateralized Credit Game (C > D) ( C > D)
Number of Banks Playing Morning stag hunt vis-à-vis Bank i prisoner’s delay risk process risk 012 dilemma dominant dominant (m,m) 4C 3C 2C Bank i (m,a) or (a,m) 3C+D 2C+D C+D D 1 (a,a) 2C+2D 2D+C 2D 0 1- 2(C-D) C 2C-D Note: We underline the cost associated with the strategy that is the best response of a bank given the strategy played by the other.
Result 9: In a repeated game setting with a trigger strategy, Game 16 Three-Player Game with Priced Credit the prisoner’s dilemma in the case of collateralized credit ( F > D) may turn into a stag hunt if the discount factor is sufficiently high. In the priced credit regime, the stag hunt game remains Number of Banks Playing Morning a stag hunt. Under both regimes, the likelihood of early vis-à-vis Bank i processing is increasing in the value placed on future costs. 012 (m,m) 2FF 0 Bank i (m,a) or (a,m) D+F D D (a,a) 2D 2D 2D
Note: We underline the cost associated with the strategy that is the best 10. More Players response of a bank given the strategy played by the other.
The number of participants in RTGS systems around the world varies significantly. In the United Kingdom, the CHAPS Sterling system has fifteen direct participants, whereas the payment flows become intertwined with the strategic Federal Reserve’s Fedwire Funds Service has more than 7,000. interaction. Understanding the network structure of payment Obviously, our two-player framework is a simplification of flows is important when analyzing behavior in interbank reality. Adding additional banks to the mix increases the payments systems (see, for example, Soramäki et al. [2007] dimensionality of the game. With three banks each having a and Bech and Garratt [2006]). We summarize the effects of dollar to send to one another, the number of different strategy expanding the number of players as: profiles increases from four to sixty-four as banks now can delay to one bank while sending early to the other. With four Result 10: Adding more players does not fundamentally banks, the same number is 4,096. change the strategic interaction of the intraday liquidity Here, we focus on the three-player game, where each bank management game. has a dollar to send to the other two banks. The settlement costs of Bank i ∈ {A, B, C}, given its own strategy and the number of other banks sending payments early to the bank in question, are shown in Game 15 and Game 16, respectively. 11. More Actions In the collateralized credit game, it is still a dominating strategy to delay. Hence, ((a, a), (a, a), (a, a)) is a unique A trend among RTGS systems has been to extend operating hours. equilibrium, but the outcome is inefficient. In the priced credit Since 2001, the Fedwire Funds Service has opened at 9:00 p.m. ET 12 game, the best response of Bank i is to do the same if the two on the preceding calendar day and closed at 6:30 p.m. ET. In opponents either send or delay all their payments. If one bank comparison, the Swiss Interbank Payment system opens at is playing morning and the other is playing afternoon, then it is 5:00 p.m. on the preceding calendar day and closes at 4:15 p.m., the best response of Bank i also to send one early and delay the thus approaching around-the-clock processing. This trend, other. However, such strategy profiles are only equilibria if the coupled with the fact that RTGS systems operate in continuous payment flow somehow miraculously forms a cycle—for 12 For example, on a Sunday, the Fedwire Funds Service will open at 9:00 p.m. example, Bank A sends to Bank B, which sends to Bank C, ET with a cycle date of Monday, although transfers sent from 9:00 p.m. to which sends to Bank A. In other words, the underlying midnight ET on Sunday will settle in real time on Sunday.
20 A Tale of Games Banks Play
Game 17 systems, include times at which ancillary payments or securities Collateralized Credit Game with Three Periods settlement systems settle their final positions. ( C > D) Result 11: Expanding the number of periods within the day Bank B does not fundamentally change the strategic interaction of ma e mC, C 2C, D 3C, 2D the intraday liquidity management game, but it does make Bank A aD, 2CC+D, C+DD+2C, 2D equilibrium selection more difficult. e 2D, 3C 2D, D+2C 2D+C, 2D+C
Note: We underline the cost associated with the strategy that is the best response of a bank given the strategy played by the other. 12. Conclusion
Game 18 This article presents a simple game-theoretical framework that Priced Credit Game with Three Periods can be used to address both positive and normative economic ( F > D) questions associated with intraday liquidity management. The Bank B simplicity of the framework is both its strength and its ma eweakness. The strength is that it clearly exposes the m 0, 0 F, D 2F, 2D fundamental trade-offs associated with strategic interaction in Bank A aD, FD, D F + D, 2D an RTGS environment. However, the extensions discussed e 2D, 2F 2D, F + D 2D, 2D highlight the complexity faced by banks in managing intraday Note: We underline the cost associated with the strategy that is the best liquidity, the challenges faced by policymakers, and response of a bank given the strategy played by the other. consequently the difficulty in devising an all-encompassing framework. Nonetheless, our analysis shows the commonality of issues faced by all stakeholders in the world’s interbank time, suggests that a model with only a morning and an payments systems. afternoon period is perhaps too coarse a representation. The ongoing relevance of the issues discussed in this article As a first step, we extend our model with an extra period is exemplified by the Federal Reserve Board’s February 28, denoted evening. In the case of collateralized credit, the prisoner’s 2008, request for public comments on proposed changes to its dilemma remains if the opportunity cost of collateral is larger than Payments System Risk policy that are intended to loosen the cost of delay. Banks end up playing (evening, evening), even intraday liquidity constraints and reduce operational risks in though it would be better for them to play either (morning, financial markets and the payments system.14 The Board is morning) or (afternoon, afternoon), as shown in Game 17. proposing a new strategy for providing intraday credit to In the case of priced credit, adding an extra period yields the depository institutions and would encourage these institutions game shown in Game 18. The game remains a coordination game to collateralize their daylight overdrafts. Specifically, the Board if the cost of liquidity is larger than the cost of delay. The strategy proposes to adopt a policy of supplying intraday balances to profile (evening, evening) is now an additional feasible Nash depository institutions predominantly through voluntarily equilibrium. As the number of periods increases, it might become collateralized daylight overdrafts. The proposed policy would increasingly difficult for banks to coordinate. In such cases, focal encourage the voluntary pledging of collateral to cover daylight points (often referred to as Schelling points), which are solutions overdrafts by providing collateralized daylight overdrafts at a that for some reason seem natural or special, may offer guidance zero fee and by raising the fee for uncollateralized daylight 13 in terms of equilibrium selection. As discussed by McAndrews overdrafts to 50 basis points (annual rate) from the current 36 and Rajan (2000), focal points may, in the context of RTGS basis points. The Board expects that a revised Payments System Risk policy could be implemented approximately two years 13 Thomas C. Schelling was awarded the Nobel Prize in Economics in 2005 “for having enhanced our understanding of conflict and cooperation through from the adoption of a final rule. game-theory analysis.” One contribution was the notion of focal points. Schelling found that coordinative solutions—which he called focal points— could be arrived at more often than predicted by theory. The ability to coordinate appears to be related to the parties’ common frames of reference. Social conventions and norms are integral parts of this common ground (see 14 See
FRBNY Economic Policy Review / September 2008 21 Appendix
Expected Settlement of Mixed Strategy Credit risk: cA()mm, = δ 20 +++δ ()1 – δ 01()–δ δ 01()– δ 20 2 Ec[]= p cmm(), ++p ()1 – p cma(), ()1 – p pc() a, m = 0 ()2 (), A 2 2 + 1 – p caa c ()ma, = δ 0 ++δ ()1 – δ 01()– δ δ ()F +θ +()1 – δ F 2 = ()1 –δ F +δ ()1 –δ θ ⎛⎞D D⎛⎞D ⎛⎞D D = ---- ρ ++---- 1 – ---- ρ 1 – ------D A 2 2 ⎝⎠ρ ρ ⎝⎠ρ ⎝⎠ρ ρ c ()a, m = δ 0 ++δ ()1 – δ 01()– δ δ D +()1 – δ D ⎛⎞D 2 = ()1 – δ D + 1 – ---- ()ρ + D ⎝⎠ρ cA()aa, = δ 20 ++δ ()1– δ 01()– δ δ D +()1 – δ 2D = ρ = ()1 –δ D
Settlement Risk Repeated Collateralized Collateralized credit liquidity risk: Credit Game cA()mm, = ω2()2CD+ ++ω ()1 –ω ()CD+ ()1 – ω ω 2C + ()1–ω 2C When is it better to play “always delay” given that the opponent = ()1 +ω C +ω D plays “trigger”? cA()ma, = ω2()2CD+ ++ω ()1 –ω ()2CD+ ()1 –ω ω 2C 2 β() + ()1 – ω 2C ------C - > ------CD+ ⇒ > ------1 - ⇒β < D β D + β C β D 1 – ---- = 2C +ω D 1 – 1 – 1 – C cA()am, = ω2()CD+ ++ω ()1 – ω D ()1 – ω ω ()CD+ 2 + ()1– ω D When is it better to play “always delay” given that the opponent = ω CD+ also plays “always delay”? cA()aa, = (ω 2 +ω()1– ω C ++()1 – ω ω C ()1 – ω 2 )()CD+ β()CD+ ()CD+ β – 1 2C + ------> ------⇒ 2C+ ------()CD+ > 0 ⇒CD> Credit risk: 1 – β 1 – β 1 – β cA()mm, = δ 2C +++δ ()1 –δ C ()1 –δ δ C ()1 –δ 2C = C When is (trigger, trigger) the risk-dominant equilibrium? cA()ma, = δ 2C ++δ ()1 – δ C ()1 – δ δ()2C + θ +()1 – δ 22C C β()CD+ β()CD+ CD+ = ()2– δ C + δ()1– δ θ ------++2C ------< D + ------+ ------⇒ A 2 2 1 – β 1 – β 1 – β 1 – β c ()am, = δ 0 ++δ ()1 – δ 01()– δ δ D +()1 – δ D 2 – β 2()CD– = ()1 – δ D 2CD< + ------D - = ------D⇒β > ------A 2 2 1 – β 1 – β 2CD– c ()aa, = δ 0++δ ()1 – δ 01()– δ δ ()CD+ +()1 – δ ()CD+ 1⎛⎞β D 2 – β = ()1 – δ ()CD+ --- F + ------D- < ------⇒ F < ------D 2⎝⎠1 – β 1 – β 1 – β Priced credit liquidity risk: cA()mm, = ω 2D +++ω()1 – ω D ()1–ω ω F ()1 –ω 20 = ω ()1 – ω F+ ω D cA()ma, = ω 2D +++ω()1 – ω D ()1 – ω ω F ()1 – ω 2F = ()1 – ω F +ω D cA()am, = ω 2D +++ω()1 – ω D ()1 – ω ω D ()1 – ω 2D = D cA()aa, = ω 2D +++ω ()1 – ω D ()1 – ω ω D ()1 – ω 2D = D
22 A Tale of Games Banks Play References
Aumann, R. 1974. “Subjectivity and Correlation in Randomized Furfine, C. H., and J. Stehm. 1998. “Analyzing Alternative Intraday Strategies.” Journal of Mathematical Economics 1, no. 1 Credit Policies in Real-Time Gross Settlement Systems.” Journal (March): 67-96. of Money, Credit, and Banking 30, no. 4 (November): 832-48.
Bech, M., and R. Garratt. 2003. “The Intraday Liquidity Greenspan, A. 1996. “Remarks on Evolving Payment System Issues.” Management Game.” Journal of Economic Theory 109, Journal of Money, Credit, and Banking 28, no. 4, part 2 no. 2 (April): 198-219. (November): 689-95.
———. 2006. “Illiquidity in the Interbank Payment System Heller, D., T. Nellen, and A. Sturm. 2000. “The Swiss Interbank Following Wide-Scale Disruptions.” Federal Reserve Bank Clearing System.” Swiss National Bank, Payment System of New York Staff Reports, no. 239, March. Subsection, June. Available at
Bech, M. L., C. Preisig, and K. Soramäki. 2008. “Global Trends in Martin, A. 2004. “Optimal Pricing of Intra-Day Liquidity.” Journal Large-Value Payments.” Federal Reserve Bank of New York of Monetary Economics 51, no. 2 (March): 401-24. Economic Policy Review 14, no.2 (September): 59-81. Martin, A., and J. McAndrews. 2008. “An Economic Analysis of Liquidity- Berger, A. N., D. Hancock, and J. C. Marquardt. 1996. “A Framework Saving Mechanisms.” Federal Reserve Bank of New York Economic for Analyzing Efficiency, Risks, Costs, and Innovations in Policy Review 14, no. 2, (September): 25-39. the Payments System.” Journal of Money, Credit, and Banking 28, no. 4, part 2 (November): 696-732. McAndrews, J., and S. Rajan. 2000. “The Timing and Funding of Fedwire Funds Transfers.” Federal Reserve Bank of New York Board of Governors of the Federal Reserve System. 2006. “Consultation Economic Policy Review 6, no. 2 (July): 17-32. Paper on Intraday Liquidity Management and Payments System Risk Policy.” Federal Register 71, no. 119: 35679-87. Docket Mills, D. C., and T. D. Nesmith. 2008. “Risk and Concentration in no. OP-1257. Payment and Securities Settlement Systems.” Journal of Monetary Economics 55, no. 3 (April): 542-53. Danmarks Nationalbank. 2003. “SCP: Scandinavian Cash Pool.” Monetary Review, fourth quarter: 23-5. Reserve Bank of New Zealand. 2006. “Reform of the Reserve Bank of New Zealand’s Liquidity Management Operations.” June. Farrell, J., and M. Rabin. 1996. “Cheap Talk.” Journal of Economic Perspectives 10, no. 3 (summer): 103-18. Skyrms, B. 2004. The Stag Hunt and the Evolution of Social Structure. Cambridge: Cambridge University Press. Freeman, S. 1996. “The Payment System, Liquidity, and Rediscounting.” American Economic Review 86, no. 5 Soramäki, K., M. L. Bech, J. Arnold, R. J. Glass, and W. E. Beyeler. 2007. (December): 1126-38. “The Topology of Interbank Payment Flows.” Physica A 379, no. 1 (June): 317-33.
FRBNY Economic Policy Review / September 2008 23
Antoine Martin and James McAndrews
An Economic Analysis of Liquidity-Saving Mechanisms
• Liquidity-saving mechanisms (LSMs) are 1.Introduction queuing arrangements for payments that operate alongside traditional real-time arge-value payments systems, used by banks to settle gross settlement (RTGS) systems. L financial and commercial transactions, play a key role in the financial system. The importance of these payments • LSMs allow banks to condition the release of systems can be illustrated by the large amounts they settle. queued payments on the receipt of offsetting Every year in the United States, the systems process value equal or partially offsetting payments; as a result, to approximately 100 times GDP. banks are less inclined to delay the sending Innovations in the design of large-value payments systems of payments. have led to many improvements in their operations. For example, over the last twenty years, many countries have • An analysis of LSMs finds that these adopted real-time gross settlement (RTGS) systems for their large-value payments. In an RTGS system, each payment is mechanisms typically perform better than settled individually, on a gross basis, at the time the payment is pure RTGS systems when it comes to sent. RTGS systems offer many advantages—for instance, they settling payments early. limit the risk exposure of payments system participants and allow for rapid final settlement of payments during the day. • RTGS systems can sometimes be preferable However, RTGS systems require large amounts of central bank to LSMs, such as when many banks that balances to function smoothly. send payments early in RTGS choose to More recent innovations have occurred in the design and queue their payments when an LSM implementation of various liquidity-saving mechanisms is available. (LSMs) that are used in conjunction with RTGS systems.1 An LSM gives participants in the payments system an additional option not offered by RTGS alone: A payment can be put into a queue and then released from the queue if some prespecified event occurs. Such mechanisms can reduce the amount of central bank balances necessary to operate the system smoothly as well as quicken the settlement of payments.
Antoine Martin is a research officer and James McAndrews The authors thank Enghin Atalay for excellent research assistance. The views a senior vice president at the Federal Reserve Bank of New York. expressed are those of the authors and do not necessarily reflect the position
FRBNY Economic Policy Review / September 2008 25 As we describe in detail below, an LSM allows banks to send the RTGS channel). If the bank submits the payment to the LSM payments conditional on the receipt of payments, and it can channel, that payment will settle only when certain conditions accommodate some netting of payments. have been met. If the bank attempts to settle via the RTGS Over the past decade, researchers have been able to simulate channel and it has sufficient funds available, the payment the performance of various LSMs. In most of these simulations, will settle immediately. the researcher makes assumptions about the behavior of the One condition that might trigger the settlement of a parties in the system and measures various consequences of the payment (and is common to both types of LSMs) occurs when assumed behavior. This approach has great potential to yield the request in the LSM channel is made in the presence of an useful answers to a number of questions. offsetting payment in the queue of the bank to which the This article outlines a different approach to the study of payment is to be made. If the two payments offset, then both LSMs in a payments system. It examines a theoretical model can be released by the LSM. of the behavior of parties, which for simplicity we refer to as A condition of another sort is determined by the type of banks. Each bank has particular motivations and constraints; LSM used. In a balance-reactive LSM, a bank has the choice of as a result, its behavior can be determined as an equilibrium reserving some level of its account balances for the settlement of payments submitted via the RTGS channel. If the bank’s balances exceed the predetermined reserve level, then that Every year in the United States, [large- bank’s payments that were previously placed in the LSM can value payments] systems process value be released. In a receipt-reactive LSM, a queued payment can be released against recent receipts of the bank (rather equal to approximately 100 times GDP. than against its accumulated balances) and at any level of balances for the bank. The research reported here reveals two sources of potential outcome in response to the incentives it faces. The theoretical value to implementing an LSM. First, an LSM gives a bank a approach has the advantage of allowing banks’ reactions to new tool not available in an RTGS system: It gives the bank alternative payments system designs to be determined within the option of making a payment conditional on the receipt a theoretical model, rather than be assumed by the researcher. of another payment. In RTGS, banks can find themselves in This approach also allows outcomes to be compared the positions of those two overly polite gentlemen in the old consistently across a number of designs. vaudeville routine, who repeatedly say to each another: “After Innovations in LSMs are numerous and, in some cases, you, Alphonse.” “No, after you, Gaston.” 2 That is, so long as quite complex. A 2005 report by the Bank for International central bank balances are costly, each bank would prefer to Settlements, “New Developments in Large-Value Payment have its counterparty make a payment first. Systems,” is an authoritative source on the many alternative In an RTGS system, there is no way to condition the systems introduced until that time. The research presented settlement of a payment upon the future receipt of a payment; here generalizes the various types of LSMs by placing their however, this course of action is possible with an LSM. The essential characteristics into one of two categories: balance LSM essentially allows the two banks to solve the precedence reactive or receipt reactive. problem that bedevils them in RTGS.3 The ability to condition A simplified description of an LSM’s operation is as follows. payments on the receipt of offsetting payments provides banks A bank wishes to make a payment and has a choice of when with some insurance against the risk of having to borrow funds to submit it to the payments system. Upon submitting the from the central bank (see, for example, Mills and Nesmith payment, the bank has a second choice to make: It can either [2008]). In turn, this means that banks are more willing to submit the payment to a central queue (the LSM part of the submit payments to the LSM earlier than they would to the payments system, which will be called the LSM channel), RTGS, which usually, but not always, has beneficial effects. or attempt to settle the payment at the time of submission (the RTGS part of the payments system, which will be called 2 The American version of this routine is based on a comic strip by Frederick Burr Opper, “Alphonse and Gaston,” which was popular in the early 1900s and 1 In this article, liquidity-saving mechanism refers to a mechanism intended to pokes fun at exaggerated politeness. economize on the use of central bank reserves. These reserves can typically be 3 Note that an internal queue, rather than an LSM, could allow a bank to send obtained intraday from the central bank either against collateral or for a small a payment conditional on receiving another payment. However, investment fee. A liquidity-saving mechanism can allow payments to be settled with fewer in an internal queue has benefits to other banks that cannot be appropriated central bank reserves. As we discuss, under some circumstances, an LSM can by the investing bank. In addition, an internal queue does not solve the indeed economize on the use of central bank balances as well as lead banks to precedence problem; the bank with the internal queue arranges for its submit payments earlier to the payments system. In general, this outcome can payments to follow the receipt from other banks. As a result, RTGS systems be defined as making the payments system more “liquid.” that have internal queues can still suffer from excessive delay in payments.
26 An Economic Analysis of Liquidity-Saving Mechanisms The second source of potential value to implementing an Our study proceeds as follows. In the next section, we LSM comes from the offsetting of payments within the queue, describe the environment in place when banks decide on their which reduces the need for central bank balances. The benefit payment submission strategies. In subsequent sections, we arising from payments offsetting in the queue is well consider various scenarios, including banks’ behavior in an understood. An important contribution of our research is RTGS system, a balance-reactive LSM, and a receipt-reactive to show that even when no such offsetting occurs, the first LSM. potential benefit described above means that an LSM can improve welfare. Several authors have examined theoretical behavior in RTGS systems. Angelini (1998, 2000) considers the behavior 2. The Environment of banks in an RTGS system in which they face delayed costs for payments as well as costly borrowing of funds. He shows that This section describes the economic environment in which the natural payment timing equilibria of RTGS systems (in the banks operate. It specifies the economic agents, which we refer absence of LSMs) involve excessive delay of payments, as banks to as banks; the banks’ objectives; and other factors that do not properly internalize the benefits to other banks from the influence banks’ decisions. We study a simple payments receipt of funds. Bech and Garratt (2003) carefully specify a system. While some features of our model are unrealistic game-theoretic environment in which they find that RTGS because of their simplicity, we believe that the model captures systems can be characterized by multiple equilibria, some of essential economic frictions that affect banks’ behavior in the which can involve excessive delay. Roberds (1999) compares payments system. The simplicity of the model allows us to gross and net payments systems with systems offering an LSM. obtain explicit results and to provide transparent intuition Examining the incentives banks have to engage in more risk- for these results.4 In our model, the essential features of an LSM are as follows. Payment orders are put in a queue, and the release of a payment The research presented here generalizes order from the queue will occur when some conditions have been satisfied. Here, the conditions are that a bank has the various types of [liquidity-saving sufficient funds and that an offsetting payment has been mechanisms] by placing their essential received. While actual LSMs can be considerably more characteristics into one of two categories: complicated, they all share those basic, essential features. By capturing the essential features shared by LSMs, our model can balance reactive or receipt reactive. describe the basic economic mechanisms associated with these systems in a tractable way. To incorporate additional features of queues, we would taking behavior in the different systems, Roberds finds that, require a more complicated model that may be difficult to solve under certain circumstances, the risk profiles of LSMs and net explicitly. Those additional features could include: 1) limits on systems are identical. Mills and Nesmith (2008) also study the the amount of central bank balances that can be committed to impact of incentives on banks’ payment patterns. McAndrews a particular bank or set of banks, 2) a time before or after which and Trundle (2001) and the Bank for International Settlements a payment should be sent, and 3) different payment priorities (2005) provide extensive that can change the ordering of payments submitted to the descriptive material on LSMs. queue away from first in, first out. Also, queues may or may not Willison (2005) examines the behavior of banks in an LSM, be transparent. If a queue is transparent, banks can see pending and this work is most similar to ours. He models agents as payments in their favor.5 In addition to the tractability issue, having an ordering of payment priority, which is similar in it is not clear that adding these features would modify our key spirit to our assumption that some banks’ payments are time insights into the potential benefits of the LSM’s essential sensitive. Willison models the extension of credit from the feature: to allow conditionality in the settlement of some central bank as an ex ante amount to be borrowed by banks, payments. while in our study the credit is tapped ex post, depending on a bank’s per-period balance. Our model extends Willison’s 4 Our analysis summarizes our more technical paper (Martin and McAndrews analysis in two dimensions that prove to be important: We 2008). 5 consider a wider array of LSMs and, crucially, we allow for In our model, queues are opaque—that is, banks cannot see particular payments in other banks’ queues. However, banks are forward looking and, in liquidity shocks, which we define as shocks to the level of equilibrium, they will expect banks of different types to have (or have not) a bank’s balances on account. submitted some payments to the LSM, and these expectations will be fulfilled.
FRBNY Economic Policy Review / September 2008 27 Each day, the payments system operates during two periods: available for free but only against collateral. Evidence provided the morning and the afternoon. A large number of banks are by the Bank for International Settlements (2005) suggests that involved, and each bank must send payments to other banks the cost of central bank balances can influence banks’ payment and will receive payments from other banks. For simplicity, behavior. all payments are of the same size. Some banks must make time-sensitive payments. If a time- We assume that banks are risk neutral. Our definition of risk sensitive payment is not settled until the afternoon, the sending neutrality is that a bank would be willing to pay up to 50 cents bank incurs a delay cost. This cost could arise because delay for the opportunity to participate in a lottery that promises one creates a bad reputation for the bank in the eyes of its dollar with probability 1/2 and nothing with probability 1/2. customers or counterparties. The behavior of banks suggests We believe that our results would extend to more complicated that they perceive some payments as being more time sensitive economies in which there are many periods and payments of than others. For example, payments made to close a real estate different sizes. Banks have rational expectations about the transaction may be more time sensitive than payments for probability of receiving a morning payment. In other words, previous deliveries of supplies to a manufacturer.7 However, banks are able to calculate the correct probability of receiving delaying non-time-critical payments until the afternoon is a payment in the morning. costless. Banks must choose whether to send their payments to Banks may receive a positive liquidity shock, a negative another bank in the morning or in the afternoon. Three factors liquidity shock, or no liquidity shock. The liquidity shock is influence this decision. First, banks face a cost if they must modeled by assuming that each bank must send a second borrow funds at the central bank. Second, some banks must payment to (and will receive a payment from) a settlement institution. For the purpose of this article, U.S. settlement institutions are represented jointly by CLS Bank, CHIPS, 8 If a time-sensitive payment is not settled and DTC. The settlement institution is intended to capture other payment or settlement systems whose pay-ins and until the afternoon, the sending bank payouts are made using the central bank’s payments system. incurs a delay cost. This cost could arise This rearrangement of balances, which we refer to as a shock, because delay creates a bad reputation introduces another source of heterogeneity between banks and leads to a more diverse set of strategic interactions. for the bank in the eyes of its customers In the model, banks find out when they must send a or counterparties. payment to the settlement institutions and when they will receive an offsetting payment from the settlement institutions before they decide whether to send their payment to other make time-sensitive payments. Third, banks may receive a banks. We assume that payments to settlement systems cannot positive liquidity shock, a negative liquidity shock, or no shock. be delayed. We say that a bank that receives a payment from the Each factor is explained in more detail below. settlement institutions in the morning and does not have to Banks face a cost if they must borrow funds from the central make an offsetting payment until the afternoon has received bank. All banks start the day with zero reserves at the central a positive liquidity shock, as its account balances are boosted bank and, because of the symmetry assumed by the model, end throughout the day prior to its late offsetting payment. the day with zero reserves as well. Banks are not allowed to hold Conversely, a bank that receives a payment from the settlement negative reserves at the end of the morning period, but they can institutions in the afternoon but must make a payment in the borrow reserves from the central bank at a cost.6 For example, 7 Armantier, Arnold, and McAndrews (2008) find that Fedwire payments tend a bank that sends a payment but does not receive a payment in to settle earlier in the day on days with higher values of customer payments. the morning period would have to borrow reserves. The cost of This behavior could reflect higher costs of delay for customer payments relative borrowing can represent either an explicit fee imposed by the to interbank payments. 8 central bank, as in the United States, or the implicit cost of CLS Bank is a payment-versus-payment settlement system that settles foreign exchange transactions in fifteen currencies. CLS Bank is operated by CLS Bank collateral, as in many countries where central bank balances are International, a bank-owned Edge Act corporation incorporated in the United States. CHIPS (the Clearing House Interbank Payments System) is a private, 6 We assume that excess reserves have zero return. In our model, banks cannot large-value U.S. dollar payments system owned and operated by the Clearing trade with one another the funds needed to make payments. This assumption House Payments Company. DTC (the Depository Trust Company) is a is a common one in the literature and reflects the empirical observation that securities settlement system that settles the majority of U.S. corporate securities markets for intraday funds are either nonexistent or operate only in and commercial paper. DTC is a wholly owned subsidiary of Depository Trust extraordinary circumstances. and Clearing Corporation.
28 An Economic Analysis of Liquidity-Saving Mechanisms morning receives a negative shock, as its balances are depleted all at the same time in the morning. Introducing the liquidity early. When the payment from the settlement institutions is shock, together with the borrowing cost, provides an incentive received in the same period in which the bank’s outgoing for banks to delay payments. For example, a bank that payment to the settlement institutions must be sent, the bank experiences a negative liquidity shock and that must make a experiences no liquidity shock, as the offsetting payments have non-time-critical payment prefers to send it in the afternoon a neutral effect on its balances. even if all other banks pay early. Indeed, in that case, the bank We assume that payments sent among banks are larger than incurs no cost if it delays its payment, but it receives the benefit payments sent to and received from settlement institutions. of not having to borrow from the central bank. This implies that while banks may face large liquidity shocks, In summary, each bank can have one of six possible profiles: these shocks are relatively small compared with the banks’ It may or may not have to make a time-sensitive payment, and may experience a positive, negative, or no liquidity shock. Exhibit 1 displays the six profiles in the form of a color key that Each bank can have one of six possible will describe the equilibria of our economy. We assign the color light blue to banks that receive a positive liquidity shock, gray profiles: It may or may not have to make to banks that receive no liquidity shock, and dark blue to banks a time-sensitive payment, and it may that receive a negative liquidity shock. A black border indicates experience a positive, negative, or no banks that must make time-sensitive payments, while a white border indicates banks that make non-time-sensitive liquidity shock. payments. The remainder of this article examines different settlement systems for the payments system described above. We consider expected payment activity. In addition, payments made real-time gross settlement alone and an RTGS system through settlement institutions occur either at the very supplemented by two types of liquidity-saving mechanisms. beginning or very end of the business day. Thus, a bank that In each case, we describe the timing of payments predicted by begins its business day with a negative liquidity shock, then the model in equilibrium. In equilibrium, every bank chooses receives a payment from another bank in the morning, and a submission time for its direct payments that minimizes its then delays sending a payment to another bank will have a delay and borrowing costs given the payment submission positive balance at the end of the morning period. In contrast, strategies of all other banks in the payments system. We also a bank that receives a positive liquidity shock at the start of the compare the desirability of different settlement systems business day, then proceeds to send a payment to another bank according to our model. in the morning, and then does not receive a payment from another bank during the morning period will end the morning with a negative balance. The latter bank must borrow from the central bank, incurring a borrowing cost in the process. With all those factors explained, we can now trace out the Exhibit 1 intuition of the model. By itself, the cost of borrowing gives Banks Using Large-Value Payments Systems banks an incentive to send payments at the same time as other banks (either in the morning or in the afternoon). For example, Positive No Negative absent liquidity shocks and delay costs, banks would be willing liquidity liquidity liquidity shock shock shock to send their payments in the morning (or afternoon) if they knew that all other banks would send theirs in the morning Time-sensitive payments (or afternoon). That way, they would never have to borrow from the central bank. However, the cost of delay provides Non-time-sensitive payments an incentive for banks to make payments early. For example, if the cost of delay is sufficiently high, a bank will prefer to send a time-critical payment early even if all other banks delay their Note: The key applies to Charts 1-10. It describes six types of banks: payments. This will happen if the cost of delay is greater than those sending time-sensitive payments that experience a positive liquidity shock, no liquidity shock, or a negative liquidity shock; and those sending the cost of borrowing. A cost of borrowing and a high cost of non-time-sensitive payments that experience a positive liquidity shock, delay will therefore result in banks preferring to send payments no liquidity shock, or a negative liquidity shock.
FRBNY Economic Policy Review / September 2008 29 3. Real-Time Gross Settlement count on offsetting payments during the morning period. In sum, in a world with liquidity shocks, we expect that all This section studies our model economy in terms of a real-time non-time-sensitive payments will always be delayed. gross settlement system. With RTGS, payments are settled on We present the equilibria associated with RTGS in Charts 1-4. an individual basis at the time they are sent. We assume that The dark gray bars indicate payments sent to settlement banks have sufficient borrowing capacity at the central bank to institutions. These payments are made either early in the make payments, even if they do not receive a prior payment morning or late in the afternoon. Payments to other banks are and/or have experienced a negative liquidity shock. Banks have indicated in colors that correspond to a specific type of sender the choice of sending payments in the morning or delaying as described in Exhibit 1. Payments sent by banks that have them until the afternoon. They compare the expected cost experienced a negative liquidity shock are indicated in dark of each option and choose the least expensive timing strategy. blue, those sent by banks that have received no liquidity shock As shown in Exhibit 2, there is a pattern to how events unfold are in gray, and those sent by banks that have experienced during the day. First, banks learn if they must send a time- a positive liquidity shock are in light blue. A dark border critical payment and if they have a positive, negative, or no corresponds to time-sensitive payments while a white border liquidity shock. Next, early payments to settlement institutions corresponds to non-time-sensitive payments. Given this color are made. Afterward, banks decide whether to send their direct scheme, the behavior of each of the six types of banks can be payments in the morning period or to delay sending them until seen within the bar, indicating whether a bank of a certain type the afternoon period. At the end of the morning period, banks sent its payments in the morning or the afternoon. incur a delay cost if they have delayed a time-sensitive payment Charts 1-4 show how the pattern of payments changes as the or a borrowing cost if they must borrow from the central bank. cost of delay decreases. In Chart 1, the cost of delay is high, so In the afternoon, delayed payments to other banks are made, all time-sensitive payments are sent early. If the cost of delay as are late payments to settlement institutions. decreases a little (Chart 2), banks that have experienced a Liquidity shocks force some banks to start the day with negative liquidity shock will choose to delay their time- negative balances. If those banks do not have time-sensitive sensitive payments. These banks have the highest expected cost payments to make, then they can be counted on to delay their of borrowing from the central bank, so if the cost of delay is not direct payments, since otherwise even the receipt of an too high, banks will aim to reduce their need to borrow from offsetting payment would mean that they must borrow. So the central bank by delaying payments. If the cost of delay banks with negative liquidity shocks and non-time-sensitive decreases still further (Chart 3), only banks that have payments will surely delay their payments until the afternoon experienced a positive liquidity shock will send their time- period. Following domino-type logic, banks—even those that critical payments early. Finally, if the cost of delay is sufficiently enjoyed a positive liquidity shock—will delay all other non- low (Chart 4), all banks will delay payments regardless of their time-sensitive payments until the afternoon, as they cannot liquidity profiles.
Exhibit 2 Real-Time Gross Settlement Timeline
Banks find out when payments are sent to and received from Banks decide whether settlement institutions (SIs) to send payments in and whether core payments the morning or to All remaining core are time-sensitive delay payments payments made
Morning period Afternoon period
Early payments Final payments to SIs made Borrowing costs incurred if to SIs made balance is negative, delay costs incurred if time-sensitive payments not made
30 An Economic Analysis of Liquidity-Saving Mechanisms Chart 1 Chart 3 Real-Time Gross Settlement with High Costs Real-Time Gross Settlement with Lower Costs of Delay of Delay
Value of payments (normalized to total 1) Value of payments (normalized to total 1) 0.6 0.8 0.7 0.5 0.6 0.4 0.5
0.3 0.4
0.3 0.2 0.2 0.1 0.1
0 0 Settlement Early entry Morning Afternoon End-of-day Settlement Early entry Morning Afternoon End-of-day system into payments settlement system into payments settlement activity LSM queue settled activity activity LSM queue settled activity
Source: Authors’ calculations. Source: Authors’ calculations. Notes: The six types of banks/colors are described in Exhibit 1. Notes: The six types of banks/colors are described in Exhibit 1. LSM is liquidity-saving mechanism. LSM is liquidity-saving mechanism.
Chart 2 Real-Time Gross Settlement with Medium Chart 4 Costs of Delay Real-Time Gross Settlement with Lowest Costs of Delay Value of payments (normalized to total 1) 0.6 Value of payments (normalized to total 1) 0.9
0.5 0.8 0.7
0.4 0.6
0.5 0.3 0.4
0.2 0.3
0.2 0.1 0.1 0 0 Settlement Early entry Morning Afternoon End-of-day Settlement Early entry Morning Afternoon End-of-day system into payments settlement system into payments settlement activity LSM queue settled activity activity LSM queue settled activity
Source: Authors’ calculations. Source: Authors’ calculations. Notes: The six types of banks/colors are described in Exhibit 1. Notes: The six types of banks/colors are described in Exhibit 1. LSM is liquidity-saving mechanism. LSM is liquidity-saving mechanism.
In general, multiple equilibria can exist. Each equilibrium If all time-sensitive payments are sent early, then the described in Charts 1-4 may arise for the same set of probability of receiving a payment in the morning is high, parameters. There can be multiple equilibria because a bank’s implying that the probability of having to borrow from the incentives to pay early depend on the behavior of other banks.9 central bank is low. If the expected cost of borrowing from the
9 By multiple equilibria, we mean that, for the same set of parameters of the central bank is low, banks have a strong incentive to make their model, different sets of beliefs are consistent with an equilibrium. time-sensitive payments early to avoid the cost of delay.
FRBNY Economic Policy Review / September 2008 31 Conversely, if all time-sensitive payments are delayed, then the are sent early, 2) only time-sensitive payments are sent early, probability of receiving a payment in the morning is low (in or 3) all payments are delayed. this case, zero). It follows that banks will have to borrow from Consider the equilibrium in which all payments are sent in the central bank if they make payments early. With a high the morning period. This is an equilibrium, regardless of the expected cost of borrowing, agents have only weak incentives magnitude of borrowing and delay costs, because if all other to send payments early, even if this would allow them to avoid banks are sending their payments in the morning, an individual the delay cost. bank is at least as well off sending its payments in the morning We can compare the welfare associated with different as delaying them until the afternoon. If the bank has a time- equilibria. Our measure of welfare is the expected utility of sensitive payment, the bank is strictly better off sending it in the the average of the six possible types of banks. To measure this morning; if the bank does not, it is equally well off sending its average payoff, we consider what would happen if all banks payments in the morning to coincide with its receipts. This type of equilibrium does not exist when some banks have experienced a negative liquidity shock. Banks have the choice of sending The costs of delay and borrowing matter for the other two equilibria. If the cost of delaying time-sensitive payments is payments in the morning or delaying them low, if the borrowing cost is high, and if banks without a time- until the afternoon. They compare the sensitive payment plan to pay in the afternoon, then banks that must make a time-sensitive payment would prefer to delay it. expected cost of each option and choose Indeed, banks want to avoid high borrowing costs, even if they the least expensive timing strategy. must incur relatively low delay costs. Likewise, if the cost of delay is high relative to the cost of borrowing, then banks with time-sensitive payments would prefer to make them early, even could meet before the beginning of the day. Before determining if banks without time-sensitive payments choose to delay their respective liquidity shocks and whether they have to make making theirs. a time-sensitive payment, all banks are identical and thus have the same preferences for payment patterns. When several equilibria coexist, welfare is highest in the equilibrium in which the greatest number of payments is sent 4. Balance-Reactive Liquidity-Saving in the early period.10 Prior to knowing its particular Mechanism situation—that is, whether it has a positive or negative balance in the morning and whether it has a time-sensitive payment— This section studies our model economy under a particular the average bank would prefer to make its payments at the same design of a liquidity-saving mechanism. An LSM can be time as all other banks. This simultaneous payment pattern thought of as a queue, into which banks enter their payments reduces banks’ expected borrowing needs. Given the level of as an alternative to sending or delaying payments outright. coordination, the average bank would prefer to make A queued payment is released in the morning if the bank that simultaneous payments in the morning, thereby reducing delay queued the payment receives an offsetting payment or if a costs. For all equilibria, our measure of welfare decreases with group of queued payments offset multilaterally. Otherwise, the cost of delay and with the cost of borrowing. the payment is released from the queue in the afternoon. We can illustrate the role liquidity shocks play in our The probability of a payment being released from the queue analysis by looking at how equilibrium outcomes change when depends on the underlying pattern of payments. Exhibits 3 and 4 these shocks are suppressed. This would correspond to the case illustrate two cases. In the exhibits, the Xs denote banks and the where all banks make their pay-ins to the settlement arrows denote a payment that must be made from one bank to institutions and receive their payouts from the settlement another. In Exhibit 3, all payments form a unique cycle that institutions early. Absent liquidity shocks, all banks start the links all banks. Here, payments in the queue cannot offset morning period with zero balances and the only difference multilaterally unless all payments are queued. In Exhibit 4, among them is whether or not their payments are time payments form cycles of length 2 that link banks in pairs. This sensitive. Three types of equilibria are possible: 1) all payments pattern of payments maximizes the probability that payments in the queue offset bilaterally. The case of a unique cycle, 10 Note that the pattern of payments such that all payments, including time- illustrated in Exhibit 3, is particularly interesting because it sensitive ones, are made early may not provide maximum welfare. This does not contradict the result stated above, as this pattern of payments is not an allows us to disentangle two different roles played by LSMs. equilibrium. One is that the LSM allows agents to condition the release of
32 An Economic Analysis of Liquidity-Saving Mechanisms Exhibit 3 a payment on the receipt of an offsetting payment. A different, A Unique Cycle though complementary, role is that an LSM can permit bilateral or multilateral netting. In the case of a unique cycle, however, no netting occurs unless all payments are in the X queue. Here we can study, in isolation, the role of the LSM in allowing banks to make payments conditional on the receipt of offsetting payments. Exhibit 5 shows the timing of payments in a system in which banks are able to submit payments to the LSM. The difference X X between this and pure RTGS is that now banks have the option of queuing their payments. In other words, a bank can condition its sending of a queued payment on the receipt of an offsetting payment, in addition to sending or delaying the X payments outright. The decision to queue a payment is made after the bank’s liquidity shock is known, allowing the bank to take into account the shock’s effect on the level of the balance before deciding whether to queue a payment. Thus, we call this Exhibit 4 a balance-reactive LSM. Later, we will consider an LSM in Cycles of Length 2 which the release of a payment from the queue does not depend on a liquidity shock. Charts 5-8 illustrate the equilibria arising with a balance- X reactive LSM. The visual coding in these charts is the same as in Charts 1-4. That is, payments from banks that experience a positive liquidity shock are indicated in light blue, those from banks that experience no liquidity shock are in gray, and those from banks that experience a negative liquidity shock are in X X dark blue. A black border indicates time-sensitive payments while a white border indicates payments that are not time sensitive. In addition, dark shading indicates payments entered in the queue, which we emphasize by adding a queue- X submission stage. Payments released from the queue have a light shading, while payments sent outright or delayed have no
Exhibit 5 Balance-Reactive Liquidity-Saving Mechanism Timeline
Banks find out when payments are sent to and received from settlement institutions (SIs) Banks decide whether and whether core payments to send, delay, or queue All remaining core are time-sensitive payments in the morning payments made
Morning period Afternoon period
Early payments Final payments to SIs made Borrowing costs incurred if to SIs made balance is negative, delay costs incurred if time-sensitive payments not made
FRBNY Economic Policy Review / September 2008 33 shading. Some payments from the queue will be released in the borrow from the central bank if they receive a payment from afternoon unless all payments are queued. another bank in the morning. Note that for parameters in In Chart 5, all payments are queued. This happens if both which the equilibria in Charts 5 and 6 coexist, the latter the cost of delay and the liquidity shocks are not too high. In equilibrium is not robust. this case, all payments are released in the morning and no delay If the cost of delay is higher still, the equilibrium presented cost is incurred. Banks with a negative liquidity shock, in Chart 7 can occur. In this equilibrium, all banks that must however, must borrow from the central bank. make time-sensitive payments choose to send them outright. If the cost of delay is higher and the liquidity shocks remain Banks that experience a negative liquidity shock delay non- moderate, the equilibrium depicted in Chart 6 can occur. In time-sensitive payments, while those that experience a positive this equilibrium, banks that experience positive liquidity or no liquidity shock queue non-time-sensitive payments. shocks choose to make time-sensitive payments outright If the cost of delay is neither very high nor very low and the during the early period. Banks that experience negative liquidity shocks are large, then the equilibrium in Chart 8 will liquidity shocks delay non-time-sensitive payments. All other occur. Because the liquidity shock is so large, banks that receive payments—time-sensitive payments from banks that a negative liquidity shock at the beginning of the day delay both experience negative or zero liquidity shocks and non-time- time-sensitive and non-time-sensitive payments. Only banks sensitive payments from banks that experience positive or zero that receive a positive liquidity shock at the beginning of the liquidity shocks—are queued. Some payments are released day choose to send time-sensitive payments early; all other from the queue in the morning and some are released in the payments are queued. afternoon. The same types of equilibria exist in a unique cycle and in The intuition is as follows. Because the cost of delay is cycles of length 2. However, the parameters under which these sufficiently high, banks that have experienced a positive equilibria exist in each case can be different. With cycles of liquidity shock prefer to insure themselves against the risk of length 2, there can be multiple equilibria, as is true with pure having to suffer the cost of delay. For this reason, they send RTGS. In contrast, with a unique cycle, only one equilibrium payments outright. Banks that receive a negative liquidity exists for any configuration of parameters. shock prefer to delay non-time-sensitive payments. By delaying To understand this difference, note that an LSM allows their own outgoing payments, these banks avoid the need to banks to condition the release of a payment on other banks’
Chart 5 Chart 6 Balance-Reactive Liquidity-Saving Mechanism (LSM) Balance-Reactive Liquidity-Saving Mechanism (LSM) with Low Costs of Delay with Medium Costs of Delay
Value of payments (normalized to total 1) Value of payments (normalized to total 1) 0.9 0.8
0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3
0.2 0.2
0.1 0.1 0 0 Settlement Early entry Morning Afternoon End-of-day Settlement Early entry Morning Afternoon End-of-day system into payments settlement system into payments settlement activity LSM queue settled activity activity LSM queue settled activity
Source: Authors’ calculations. Source: Authors’ calculations. Notes: The six types of banks/colors are described in Exhibit 1. Dashed Notes: The six types of banks/colors are described in Exhibit 1. Dashed borders indicate payments submitted to the queue. Shading indicates borders indicate payments submitted to the queue. Shading indicates payments released from the queue. An absence of shading indicates payments released from the queue. An absence of shading indicates payments either sent outright or delayed (not queued). payments either sent outright or delayed (not queued).
34 An Economic Analysis of Liquidity-Saving Mechanisms actions, which reduces strategic interaction between banks. higher with the balance-reactive LSM if the liquidity shocks In RTGS, by contrast, the level of strategic interaction is high: are small, if the cost of delay is either sufficiently high or Whether or not a bank chooses to send a payment during the sufficiently low, and if the probability of a liquidity shock early period depends in large part on that bank’s beliefs about occurring is large compared with the fraction of time-sensitive the strategy other banks have adopted for the timing of their payments. In one specific case, a balance-reactive LSM can lead payments. With the LSM, the bank need not concern itself with to a loss of some beneficial coordination and result in lower such complicated reasoning because it can simply submit a welfare. payment to settle automatically if an offsetting payment is When liquidity shocks are small, banks incur a large received, decoupling the bank’s decision to submit the borrowing cost only if they send a payment early and do not payment to the LSM from the bank’s beliefs about the plans receive an offsetting payment. An LSM increases welfare of other banks. because it allows banks to insure themselves against this risk Whenever multilateral netting is expected to occur in the because the bank’s payment is released from the queue only queue, a new set of strategic interactions emerges. A bank’s when an offsetting payment is received. Also, because banks incentives to submit payments to the queue depend on the can queue their payments rather than delay them outright, bank’s belief about what other banks will do. If many other more payments are released in the morning and fewer banks banks submit their payments to the queue, then more netting incur a delay cost. will occur and the incentive to queue will be high. In contrast, If the cost of delay is sufficiently high, banks send all time- if only a few other banks submit their payments to the queue, sensitive payments early, regardless of the settlement less netting will occur and the incentive to queue will be low. mechanism. However, banks that do not receive a negative As a result, the possibility of multilateral netting of queued liquidity shock queue their payments when an LSM is available, payments reintroduces strategic interaction into banks’ rather than delay them as they would in RTGS. Hence, more submission behavior, leading to multiple equilibria. In payments are released early and the expected cost of borrowing contrast, since there is no netting in a unique cycle, multiple is reduced for banks that send payments outright. If the cost of equilibria do not occur. delay is sufficiently low, all banks will find it beneficial to queue We can compare the welfare of banks if settlement occurs payments when an LSM is available so that all payments are according to RTGS or to a balance-reactive LSM. Welfare is released early.
Chart 7 Chart 8 Balance-Reactive Liquidity-Saving Mechanism (LSM) Balance-Reactive Liquidity-Saving Mechanism (LSM) with High Costs of Delay with Large Liquidity Shock
Value of payments (normalized to total 1) Value of payments (normalized to total 1) 0.7 0.50 0.45 0.6 0.40 0.5 0.35
0.4 0.30 0.25 0.3 0.20
0.2 0.15 0.10 0.1 0.05 0 0 Settlement Early entry Morning Afternoon End-of-day Settlement Early entry Morning Afternoon End-of-day system into payments settlement system into payments settlement activity LSM queue settled activity activity LSM queue settled activity
Source: Authors’ calculations. Source: Authors’ calculations. Notes: The six types of banks/colors are described in Exhibit 1. Dashed Notes: The six types of banks/colors are described in Exhibit 1. Dashed borders indicate payments submitted to the queue. Shading indicates borders indicate payments submitted to the queue. Shading indicates payments released from the queue. An absence of shading indicates payments released from the queue. An absence of shading indicates payments either sent outright or delayed (not queued). payments either sent outright or delayed (not queued).
FRBNY Economic Policy Review / September 2008 35 Exhibit 6 Receipt-Reactive Liquidity-Saving Mechanism Timeline
Banks find out when Banks decide whether payments are sent to and to send, delay, or queue received from settlement payments before knowing institutions (SIs) and the balance resulting All remaining core whether core payments from the SI activity payments made are time-sensitive
Morning period Afternoon period
Early payments Final payments to SIs made Borrowing costs incurred if to SIs made balance is negative, delay costs incurred if time-sensitive payments not made
If the fraction of time-sensitive payments is relatively small, payments early. In this case, the presence of a queue unravels then the benefit of queuing non-time-sensitive payments that coordination. under a balance-reactive LSM—rather than delaying them, Again, we can illustrate the role of liquidity shocks by which would occur in RTGS—is large. Here too a balance- considering what happens when they are suppressed. Absent reactive LSM provides higher welfare than would RTGS. liquidity shocks, all payments are released early. This could If none of these conditions is satisfied, an instance may arise happen if all the payments are queued, if they are all sent in the in which RTGS provides higher welfare than does a balance- morning outright, or if some are queued and the others are sent reactive LSM. There exists a set of parameters for which the in the morning outright. equilibrium described in Chart 2 is the unique RTGS equilibrium. In this equilibrium, banks that do not receive a 5. Receipt-Reactive Liquidity-Saving We can illustrate the role of liquidity Mechanism
shocks by considering what happens We now consider a different LSM design. Previously, we when they are suppressed. Absent assumed that banks could make their decision to queue a payment conditional on their liquidity shock or, equivalently, liquidity shocks, all payments are on their balance. Here we assume that banks do not know their released early. liquidity shock when they decide either to queue payments or to pay in the morning period. We also assume that the decision to queue is irrevocable. Since banks can condition their liquidity shock send a time-critical payment early. For this set behavior only on the receipt of other payments, but not on of parameters, the best LSM equilibrium is described in Chart 8. their balance, we call this a receipt-reactive LSM.11 In this equilibrium, banks that do not receive a liquidity shock This case illustrates another possible feature of an LSM that prefer to queue a time-critical payment, rather than send it can affect bank behavior. In the receipt-reactive LSM, banks are early, because it provides them with some insurance against the given a tool that enables them to commit to making a payment risk of having to borrow from the central bank. However, this at a particular time. This ability is valuable to banks. The new action lowers the number of payments that are released in the timing of events is shown in Exhibit 6.
morning and reduces welfare. This example shows that RTGS 11 Receipt-reactive LSMs were first discussed and studied in Johnson, can create beneficial coordination among banks to send some McAndrews, and Soramäki (2004).
36 An Economic Analysis of Liquidity-Saving Mechanisms Chart 9 Charts 9 and 10 illustrate the two types of equilibria that can Receipt-Reactive Liquidity-Saving Mechanism (LSM) occur with a receipt-reactive LSM. In the equilibrium described with Low Costs of Delay in Chart 9, all payments are queued and hence released in the
Value of payments (normalized to total 1) morning. Note that early payments to settlement institutions, 0.9 which represent the liquidity shock, occur after payments have 0.8 been sent to the queue. In the equilibrium described in Chart 10, all time-sensitive payments are made early and non-time- 0.7 sensitive payments are delayed. 0.6 We can now compare welfare among RTGS, a balance- 0.5 reactive LSM, and a receipt-reactive LSM. In our model, 0.4 a receipt-reactive LSM always provides welfare at least as high as RTGS. This outcome stands in contrast to a balance-reactive 0.3 LSM, which provides lower welfare than RTGS under the 0.2 circumstances outlined above. A receipt-reactive LSM may or 0.1 may not provide higher welfare than a balance-reactive LSM 0 does. If the cost of delay is sufficiently large, for example, a Early entry Settlement Morning Afternoon End-of-day balance-reactive LSM will provide higher welfare. A balance- into system payments settlement LSM queue activity settled activity reactive LSM will also provide higher welfare when the cost of delay and the probability of a liquidity shock are small. Source: Authors’ calculations. Notes: The six types of banks/colors are described in Exhibit 1. Dashed borders indicate payments submitted to the queue. Shading indicates payments released from the queue. An absence of shading indicates payments either sent outright or delayed (not queued). 6.Conclusion
This article studies a model in which banks settle daily payments while seeking to minimize the costs associated with payment delays and intraday borrowing. The novel feature of our model is that banks are subject to two types of shocks. First, banks are randomly assigned to have either time-critical payments, whose late-period settlement imposes a cost on the Chart 10 bank, or non-time-critical payments. Second, banks are subject Receipt-Reactive Liquidity-Saving Mechanism (LSM) to liquidity shocks at the start of the day because of the nature with High Costs of Delay of settlement institutions’ operations. Together, these two Value of payments (normalized to total 1) types of shocks yield a rich array of strategic situations. The 0.6 important parameters in our model are the cost of delay, the 0.5 cost of borrowing intraday funds from the central bank, the relative size of the payments made to the settlement system 0.4 with respect to bank-to-bank payments, and the proportion of time-critical payments. 0.3 To model the working of a balance-reactive liquidity-saving mechanism, we study two extreme cases. In the first, there is no 0.2 possibility of netting any strict subset of payments. In the second, payments offset each other bilaterally. These two 0.1 models provide different motives for using the LSM. In the first
0 case, banks do not assign payments to the LSM queue in the Early entry Settlement Morning Afternoon End-of-day hope of offsetting them within the queue, but rather to have into system payments settlement them settle only conditional on receiving another payment. LSM queue activity settled activity In the second, banks can also anticipate that, some of the time, Source: Authors’ calculations. their queued LSM payments will be offset and will settle inside Note: The six types of banks/colors are described in Exhibit 1. the queue.
FRBNY Economic Policy Review / September 2008 37 In most cases, the presence of a balance-reactive LSM settlement systems are not too large, the balance-reactive LSM increases welfare compared with a real-time gross settlement yields a better outcome than its receipt-reactive counterpart. As system alone, but, perhaps surprisingly, welfare may also be a result, while our results point to LSMs being at least weakly reduced. A balance-reactive LSM provides higher welfare if the preferred to RTGS for all parameter configurations, the cost of delay is high enough or low enough, and if the size of the practical choice can present more of a dilemma to the operator outside settlement system and the proportion of time-critical of the large-value payments system. The dilemma is that our payments are relatively low. When this is not true, RTGS can results show that the LSM design matters. If the wrong LSM is achieve higher welfare. The intuition is that RTGS creates some implemented, it can yield either lower welfare than RTGS or beneficial coordination of payments that can be undone by the lower welfare than a competing LSM design. The challenge for presence of a queue. In our example, some banks that send a payments system operator is to know the sizes of the four payments early under the unique RTGS equilibrium choose to parameters of interest. Here we have considered basic design put the payments in the queue when they are available. The elements in choosing the LSMs to model; more complex resulting reduction in the number of payments settled early designs would introduce other behavioral considerations that leads to lower welfare. are beyond the scope of this article. With a receipt-reactive LSM, the level of welfare achieved is Future research in this area would thus benefit from always at least as high as the level achieved in RTGS alone. Here focusing on the empirical magnitudes of the parameters of the intuition is simpler. As banks cannot condition the sending interest. The cost of delaying payments and the proportion of of payments on balances, they either submit all their payments payments that are time critical are especially important to to the queue or they simply make all the time-critical payments measure and difficult to observe. Research employing in the early period. alternative distributions of these parameters will be important, In comparing balance-reactive and receipt-reactive LSMs, as will be the extension of the current model to include several we find that when delay costs are high and the payments to periods.
38 An Economic Analysis of Liquidity-Saving Mechanisms References
Angelini, P. 1998. “An Analysis of Competitive Externalities in Gross Martin, A., and J. McAndrews. 2008. “Liquidity-Saving Mechanisms.” Settlement Systems.” Journal of Banking and Finance 22, no. 1 Journal of Monetary Economics 55, no. 3 (April): 554-67. (January): 1-18. McAndrews, J., and J. Trundle. 2001. “New Payment System Designs: ———. 2000. “Are Banks Risk Averse? Intraday Timing of Operations Causes and Consequences.” Bank of England Financial in the Interbank Market.” Journal of Money, Credit, and Stability Review 11, December: 127-36. Banking 32, no. 1 (February): 54-73. Mills, D. C., and T. D. Nesmith. 2008. “Risk and Concentration in Armantier, O., J. Arnold, and J. McAndrews. 2008. “Changes in the Payment and Securities Settlement Systems.” Journal of Timing Distribution of Fedwire Funds Transfers.” Federal Reserve Monetary Economics 55, no. 3 (April): 542-53. Bank of New York Economic Policy Review 14, no. 2 (September): 83-112. Roberds, W. 1999. “The Incentive Effects of Settlement Systems: A Comparison of Gross Settlement, Net Settlement, and Gross Bank for International Settlements. 2005. “New Developments in Settlement with Queuing.” Bank of Japan Institute for Large-Value Payment Systems.” Committee on Payment and Monetary and Economic Studies Discussion Paper Series, Settlement Systems, Publication no. 67. May. Basel, Switzerland. no. 99-E-25, September.
Bech, M. L., and R. Garratt. 2003. “The Intraday Liquidity Willison, M. 2005. “Real-Time Gross Settlement and Hybrid Payments Management Game.” Journal of Economic Theory 109, Systems: A Comparison.” Bank of England Working Paper no. 2 (April): 198-219. no. 252.
Johnson, K., J. J. McAndrews, and K. Soramäki. 2004. “Economizing on Liquidity with Deferred Settlement Mechanisms.” Federal Reserve Bank of New York Economic Policy Review 10, no. 3 (December): 51-72.
FRBNY Economic Policy Review / September 2008 39
Todd Keister, Antoine Martin, and James McAndrews
Divorcing Money from Monetary Policy
• Many central banks operate in a way that 1.Introduction creates a tight link between money and monetary policy, as the supply of reserves onetary policy has traditionally been viewed as the must be set precisely in order to implement M process by which a central bank uses its influence over the target interest rate. the supply of money to promote its economic objectives. For example, Milton Friedman (1959, p. 24) defined the tools of • Because reserves play other key roles in the monetary policy to be those “powers that enable the [Federal economy, this link can generate tensions with Reserve] System to determine the total amount of money in
central banks’ other objectives, particularly existence or to alter that amount.” In fact, the very term in periods of acute market stress. monetary policy suggests a central bank’s policy toward the supply of money or the level of some monetary aggregate. • An alternative approach to monetary policy In recent decades, however, central banks have moved away implementation can eliminate the tension from a direct focus on measures of the money supply. The between money and monetary policy primary focus of monetary policy has instead become the value by “divorcing” the quantity of reserves of a short-term interest rate. In the United States, for example, from the interest rate target. the Federal Reserve’s Federal Open Market Committee (FOMC) announces a rate that it wishes to prevail in the federal funds market, where overnight loans are made among • By paying interest on reserve balances at commercial banks. The tools of monetary policy are then used its target interest rate, a central bank can to guide the market interest rate toward the chosen target. For increase the supply of reserves without this reason, we follow the common practice of using the term driving market interest rates below the target. monetary policy to refer to a central bank’s interest rate policy. It is important to realize, however, that the quantity of • This “floor-system” approach allows the money and monetary policy remain fundamentally linked central bank to set the supply of reserve under this approach. Commercial banks hold money in the balances according to the payment or form of reserve balances at the central bank; these balances liquidity needs of financial markets while are used to meet reserve requirements and make interbank simultaneously encouraging the efficient allocation of resources.
Todd Keister and Antoine Martin are research officers and James McAndrews The views expressed are those of the authors and do not necessarily reflect a senior vice president at the Federal Reserve Bank of New York. the position of the Federal Reserve Bank of New York or the Federal Reserve
FRBNY Economic Policy Review / September 2008 41 payments. The quantity of reserve balances demanded by functioning of the payments system and financial markets banks varies inversely with the short-term interest rate because more generally. As a result of this action, the fed funds rate this rate represents the opportunity cost of holding reserves. fell substantially below the target level for several days.1 The central bank aims to manipulate the supply of reserve During the financial turmoil that began in August 2007, the balances—for example, through open market operations that tension was much longer lasting. Sharp increases in spreads exchange reserve balances for bonds—so that the marginal between the yields on liquid and illiquid assets indicated a classic liquidity shortage: an increased demand for liquid assets relative to their illiquid counterparts. By increasing the supply [The] link between money and monetary of the most liquid asset in the economy—bank reserves—the Federal Reserve could likely have eased the shortage and helped policy can generate tension with central push spreads back toward more normal levels. Doing so, banks’ other objectives because bank however, would have driven the market interest rate below the reserves play other important roles FOMC’s target rate and thus interfered with monetary policy objectives. Instead, the Federal Reserve developed new, indirect in the economy. methods of supplying liquid assets to the private sector, such as providing loans of Treasury securities against less liquid collateral through the Term Securities Lending Facility. value of a unit of reserves to the banking sector equals the target Recently, attention has turned to an alternative approach interest rate. The interbank market for short-term funds will to monetary policy implementation that has the potential to then clear with most trades taking place at or near the target eliminate the basic tension between money and monetary rate. In other words, the quantity of money (especially reserve balances) is chosen by the central bank in order to achieve its interest rate target. It is important to understand the tension This link between money and monetary policy can generate tension with central banks’ other objectives because bank between the daylight and overnight need reserves play other important roles in the economy. In for reserves and the potential problems particular, reserve balances are used to make interbank that may arise. payments; thus, they serve as the final form of settlement for a vast array of transactions. The quantity of reserves needed for payment purposes typically far exceeds the quantity consistent with the central bank’s desired interest rate. As a result, central policy by effectively “divorcing” the quantity of reserves from banks must perform a balancing act, drastically increasing the the interest rate target. The basic idea behind this approach supply of reserves during the day for payment purposes is to remove the opportunity cost to commercial banks of through the provision of daylight reserves (also called daylight holding reserve balances by paying interest on these balances credit) and then shrinking the supply back at the end of the day at the prevailing target rate. Under this system, the interest rate to be consistent with the desired market interest rate. paid on reserves forms a floor below which the market rate Recent experience has shown that central banks perform cannot fall. The supply of reserves could therefore be increased this balancing act well most of the time. Nevertheless, it is substantially without moving the short-term interest rate away important to understand the tension between the daylight and from its target. Such an increase could be used to provide overnight need for reserves and the potential problems that liquidity during times of stress or to reduce the need for may arise. One concern is that central banks typically provide daylight credit on a regular basis.2 A particular version of the daylight reserves by lending directly to banks, which may “floor-system” approach has recently been adopted by the expose the central bank to substantial credit risk. Such lending Reserve Bank of New Zealand. may also generate moral hazard problems and exacerbate the It should be noted that adopting a floor-system approach too-big-to-fail problem, whereby regulators would be reluctant requires the central bank to pay interest on reserves, something to close a financially troubled bank. 1 The tension is clearest during times of acute stress in Intraday volatility of the fed funds rate remained high, with trades being executed far from the target rate, for several weeks. See McAndrews and Potter financial markets. In the days following September 11, 2001, (2002) and Martin (forthcoming) for detailed discussions. for example, the Federal Reserve provided an unusually large 2 This approach has been advocated in various forms by Woodford (2000), quantity of reserves in order to promote the efficient Goodfriend (2002), Lacker (2006), and Whitesell (2006b).
42 Divorcing Money from Monetary Policy the Federal Reserve has historically lacked authorization to do. Exhibit 1 However, the Financial Services Regulatory Relief Act of 2006 Monetary Policy Implementation will give the Federal Reserve, for the first time, explicit authority in the United States to pay interest on reserve balances, beginning on October 1, 2011. A floor system will therefore soon be a feasible option Interest rate for monetary policy implementation in the United States. In this article, we present a simple, graphical model of the Penalty rate monetary policy implementation process to show how the floor system divorces money from monetary policy. Our aim Demand for reserves is to present the fundamental ideas in a way that is accessible to a broad audience. Section 2 describes the process by which Target rate monetary policy is currently implemented in the United States Required and in other countries. Section 3 discusses the tensions that can reserves arise in this framework between monetary policy and Reserve 0 Target balances payments/liquidity policy. Section 4 illustrates how the floor supply system works; it also discusses potential issues associated with adopting this type of system in a large economy such as the United States. Section 5 concludes. smaller than its requirement, it pays a penalty that is proportional to the shortfall. Second, banks experience unanticipated late-day payment flows into and out of their reserve account after the interbank market has closed. A bank’s 2.An Overview of Monetary Policy final reserve balance, therefore, may be either higher or lower Implementation than the quantity of reserves it chooses to hold in the interbank market. This uncertainty makes it difficult for a bank to satisfy In this section, we describe a stylized model of the process its requirement exactly and generates a “precautionary” through which many of the world’s central banks implement demand for reserves. monetary policy. Our model focuses on the relationship For simplicity, we abstract from a number of features of between the demand for reserve balances and the interest rate reality that, while important, are not essential to understanding in the interbank market for overnight loans. Following Poole the basic framework. For example, we assume that reserve (1968), a variety of papers have developed formal models of requirements must be met on a daily basis, rather than on portfolio choice by individual banks and derived the resulting average over a two-week reserve maintenance period. aggregate demand for reserves.3 Our graphical model of Alternatively, one can interpret our model as applying to aggregate reserve demand is consistent with these more formal average reserve balances (and the average overnight interest approaches. We first discuss the system currently used in the rate) over a maintenance period. In addition, we do not United States and then describe a symmetric channel system, explicitly include vault cash in the analysis, using the terms as used by a number of other central banks. reserve balances and reserves interchangeably.4 Exhibit 1 presents the aggregate demand for reserves in our framework. The horizontal axis measures the total quantity of reserve balances held by banks while the vertical axis measures 2.1 Monetary Policy Implementation the market interest rate for overnight loans of these balances. in the United States The penalty rate labeled on the vertical axis represents the interest rate a bank pays if it must borrow funds at the end of We begin by examining the total demand for reserve balances by the U.S. banking system. In our stylized framework, this 4 Required reserves should therefore be interpreted as a bank’s requirement demand is generated by a combination of two factors. First, net of its vault cash holdings. To the extent that vault cash holdings are independent of the overnight rate, at least over short horizons, including them banks face reserve requirements. If a bank’s final balance is in our model would have no effect. We also abstract from the Contractual Clearing Balance program, which allows banks to earn credit for priced services 3 Recent contributions include Furfine (2000), Guthrie and Wright (2000), at the Federal Reserve by holding a contractually agreed amount of reserves in Bartolini, Bertola, and Prati (2002), Clouse and Dow (2002), Whitesell (2006a, b), excess of their requirement; these contractual arrangements, once set, act much and Ennis and Weinberg (2007). like reserves requirements.
FRBNY Economic Policy Review / September 2008 43 the period to meet its requirement. One can interpret this penalty rate—against the possibility that it will end up holding penalty rate as the interest rate charged at the Federal Reserve’s more reserves than are required. As no interest is paid on primary credit facility (the discount window), adjusted by any reserves, holding excess balances is also costly. The resulting “stigma” costs that banks perceive to be associated with demand for reserve balances will vary inversely with the market interest rate, since this rate represents the opportunity cost of holding reserves. The less expensive it is to hold precautionary The . . . demand for reserve balances will reserve balances, the greater the quantity demanded by the banking system will be. This reasoning generates the vary inversely with the market interest downward-sloping part of the demand curve in the exhibit. rate, since this rate represents the If the market interest rate were very low—close to zero—the opportunity cost of holding reserves. opportunity cost of holding reserves would be very small. In this case, each bank would hold enough precautionary reserves to be virtually certain that unforeseen payment flows will not decrease its reserve balance below the required level. In other borrowing at this facility. The important feature of the penalty words, each bank would choose to be “fully insured” against rate is that it lies above the FOMC’s target interest rate.5 the possibility of falling short of its requirements. The point in To explain the shape of the demand curve in the exhibit, we Exhibit 1 where the demand curve intersects the horizontal axis ask: given a particular value for the interest rate, what quantity represents the total of this fully insured quantity of reserve of reserve balances would banks demand to hold if that rate balances for all banks. The banking system will not demand prevailed in the interbank market? First, note that if the market more than this quantity of reserve balances as long as there interest rate were above the penalty rate, there would be an is some opportunity cost, no matter how small, of holding arbitrage opportunity: banks could borrow reserves at the these reserves. (lower) penalty rate and lend them at the (higher) market If the market interest rate were exactly zero, however, there interest rate. If the market interest rate were exactly equal to would be no opportunity cost of holding reserves. In this the penalty rate, however, banks would be willing to hold some limiting case, there is no cost at all to a bank of holding reserve balances toward meeting their requirements. In fact, additional reserves above the fully insured amount. The each bank would be indifferent between holding reserves demand curve is therefore flat along the horizontal axis after directly and borrowing at the penalty rate as long as it is sure this point; banks are indifferent between any quantities of that late-day payment inflows will not leave it holding excess reserves above the fully insured amount when the market balances at the end of the day. As a result, the demand curve interest rate is exactly zero. is flat—reflecting this indifference—at the level of the penalty Needless to say, our model of reserve demand abstracts rate for sufficiently small levels of reserve balances. from important features of reality. Holding more reserves, For interest rates below the penalty rate, each bank will for example, might require a bank to raise more deposits and choose to hold a quantity of reserves that is close to the level of subject it to higher capital requirements. Nevertheless, the its requirement; hence, aggregate reserve demand will be close model is useful because it lays out, in perhaps the simplest way to the total level of required reserves. However, as described possible, the basic relationship between the market interest rate above, banks face uncertainty about their final account balance and the demand for reserves that results from the optimal that prevents them from being able to meet their requirement portfolio decisions of banks. Moreover, small changes in the exactly. Instead, each bank must balance the possibility of shape of the demand curve would have no material effect on falling short of its requirement—and being forced to pay the the analysis that follows. The equilibrium interest rate in our model is determined by 5 The interest rate charged on discount window loans has been set above the the height of the demand curve at the level of reserve balances FOMC’s target rate since the facility was redesigned in 2003. The gap between supplied by the Federal Reserve. If the supply is smaller than the two rates was initially set at 100 basis points, but has since been lowered the total amount of required reserves, for example, the to 50 basis points (in August 2007) and to 25 basis points (in March 2008). In addition, there is evidence that banks attach a substantial nonpecuniary cost equilibrium interest rate would be near the penalty rate. If, to borrowing from the discount window, as they sometimes borrow in the however, the supply of reserves were very large, the equilibrium interbank market at interest rates significantly higher than the discount interest rate would be zero. Between these two extremes, on window rate. These stigma costs may reflect a fear that other market participants will find out about the loan and interpret it as a sign of financial the downward-sloping portion of the demand curve, there weakness on the part of the borrowing bank. is a liquidity effect of reserve balances on the market interest
44 Divorcing Money from Monetary Policy rate: a higher supply of reserves will lower the equilibrium 2.2 Symmetric Channel Systems interest rate.6 As shown in the exhibit, there is a unique level of reserve Many central banks use what is known as a symmetric channel supply that will lead the market to clear at the FOMC’s (or corridor) system for monetary policy implementation. Such announced target rate; we call this level the target supply. systems are used, for example, by the European Central Bank Monetary policy is implemented through open market (ECB) and by the central banks of Australia, Canada, England, operations that aim to set the supply of reserves to this target and (until spring 2006) New Zealand. The key features of a level. This process requires the Fed’s Open Market Desk to symmetric channel system are standing central bank facilities accurately forecast both reserve demand and changes in the that lend to and accept deposits from commercial banks. The existing supply of reserves attributable to autonomous factors lending facility resembles the discount window in the United States; banks are permitted to borrow freely (with acceptable collateral) at an interest rate that is a fixed number of basis Monetary policy is implemented in the points above the target rate. The deposit facility allows banks to earn overnight interest on their excess reserve holdings at a United States by changing the supply of rate that is the same number of basis points below the target. reserves in such a way that the fed funds In this way, the interest rates at the two standing facilities form market will clear at the desired rate. a “channel” around the target rate. Exhibit 2 depicts the demand for reserve balances in a In other words, the stock of “money” symmetric channel system. The curve looks very similar to that is set in order to achieve a monetary in Exhibit 1. There is no demand for reserves in the interbank policy objective. market if the interest rate is higher than the rate at the lending facility.8 For lower values of the market rate, the demand is decreasing in the interest rate—and hence the liquidity effect is present—for exactly the same reasons as before. Banks choose such as payments into and out of the Treasury’s account. their reserve holdings to balance the potential costs of falling Forecasting errors will lead the actual supply to deviate from short of their requirement against the potential costs of ending the target and, hence, will cause the market interest rate to with excess reserves. When the opportunity cost of holding differ from the target rate. In our simple model, the downward- reserves is lower, banks’ precautionary demand for reserves will sloping portion of the demand curve may be quite steep, be larger. indicating that relatively small forecasting errors could lead to substantial interest rate volatility. In reality, a variety of institutional arrangements, including reserve maintenance Exhibit 2 periods, are designed to flatten this curve and thus limit the A Symmetric Channel System of Monetary Policy Implementation volatility associated with forecasting errors.7 The key point of this discussion is that monetary policy is Interest rate implemented in the United States by changing the supply of reserves in such a way that the fed funds market will clear at Lending rate Demand for the desired rate. In other words, the stock of “money” is set in reserves order to achieve a monetary policy objective. This direct Target rate relationship between money and monetary policy generates the tensions that we discuss in Section 3. Deposit rate Required reserves Reserve 0 Target balances supply
6 See Hamilton (1997), Carpenter and Demiralp (2006a), and Thornton (2006) for empirical evidence of this liquidity effect. 8 The lending facility in a channel system is typically designed in a way that aims 7 See Ennis and Keister (2008) for a detailed discussion of interest rate volatility to minimize stigma effects. For this reason, we begin the demand curve in in this basic framework. See Whitesell (2006a) for a formal model of the Exhibit 2 at the lending rate instead of at a penalty rate that includes stigma “flattening” effect of reserve maintenance periods. effects, as was the case in Exhibit 1.
FRBNY Economic Policy Review / September 2008 45 The new feature in Exhibit 2 is that the demand curve does reserves” line lies on the vertical axis. The important point not decrease all the way to the horizontal axis, but instead here, however, is that regardless of these operational details, a becomes flat at the deposit rate. In other words, the deposit symmetric channel system links the quantity of reserves to the rate forms a floor below which the demand curve will not fall. central bank’s interest rate target, exactly as in the U.S. system. If the market rate were below the deposit rate, an arbitrage opportunity would exist—a bank could borrow at the (low) market rate and earn the (higher) deposit rate on these funds, making a pure profit. The demand for reserves would be unbounded in this case; such arbitrage activity would quickly 3. Payments, Liquidity Services, drive up the market rate until it at least equals the deposit rate. and Reserves The demand curve is flat at the deposit rate for the same reason it was flat on the horizontal axis in Exhibit 1. If the The link between money and monetary policy described above market rate were exactly equal to the deposit rate, banks would can generate tension with central banks’ other objectives, particularly those regarding the payments system and the provision of liquidity. Reserve balances are useful to banks, [In a symmetric channel system,] the and to the financial system more generally, for purposes other than simply meeting reserve requirements. Banks use reserve target interest rate determines, through balances to provide valuable payment services to depositors. the demand curve, a target supply of In addition, these balances assist the financial sector in reserves, and the central bank aims to allocating other, less liquid assets. Since reserves are a universally accepted asset, they can be exchanged more easily change total reserve supply to bring for other assets than any substitute. Finally, reserve balances it as close as possible to this target. serve as a perfectly liquid, risk-free store of value, which is particularly useful during times of market turmoil. Because reserves play these other important roles, the quantity of face no opportunity cost of holding excess reserves. Holding reserve balances consistent with the central bank’s monetary additional funds on deposit and lending them would yield policy objective may at times come into conflict with the exactly the same return. Banks would therefore be indifferent quantity that is desirable for other purposes. In this section, between any quantities of reserves above the fully insured we describe some of the tensions that can arise. amount. In other words, paying interest on excess reserves raises the floor where the demand curve is flat from an interest rate of zero (as in Exhibit 1) to the deposit rate (as in Exhibit 2). 3.1 Payments Policy The equilibrium interest rate is determined exactly as before, by the height of the demand curve at the level of reserve The value of the payments made during the day in a central balances supplied by the central bank. Monetary policy is thus bank’s large-value payments system is typically far greater than implemented in much the same way as it is in the United States. the level of reserve balances held by banks overnight. (In the The target interest rate determines, through the demand curve, United States, for example, during the first quarter of 2008 the a target supply of reserves, and the central bank aims to change average daily value of transactions over the Fedwire Funds total reserve supply to bring it as close as possible to this target. Service was approximately 185 times the value of banks’ total Importantly, the link between money and monetary policy balances on deposit at the Federal Reserve.) The discrepancy remains: the quantity of reserves is set in order to achieve the has widened in recent decades as most central banks have desired interest rate. adopted a real-time gross settlement (RTGS) design for their The symmetric channel systems used by various central large-value payments system, which requires substantially banks differ in a variety of important details. The Bank of larger payment flows than earlier designs based on netting of England and the ECB operate relatively wide channels, with the payment values.9 standing facility rates 100 basis points on either side of the As a result, banks’ overnight reserve holdings are too small target. Australia and Canada, in contrast, operate narrow to allow for the smooth functioning of the payments system channels, where this figure is only 25 basis points. Australia and Canada have no required reserves; in this case, the demand 9 See Bech and Hobijn (2007) for an analysis of the adoption of RTGS systems curve in Exhibit 2 shifts to the left so that the “required by various central banks.
46 Divorcing Money from Monetary Policy during the day. When reserves are scarce or costly during the Moreover, collateralizing daylight loans simply moves the day, banks must expend resources in carefully coordinating the central bank’s claims ahead of the deposit insurance fund in the timing of their payments. If banks delay sending payments to event of a bank failure, without necessarily reducing the overall economize on scarce reserves, the risk of an operational failure risk of the consolidated public sector. or gridlock in the payments system tends to increase. The Routine daylight lending by the central bank may also create combination of limited overnight reserve balances and the moral hazard problems, leading banks to hold too little much larger daylight demand for reserves thus creates tension liquidity and, perhaps, take on too much risk. In addition, such lending might make regulators more reluctant to close a financially troubled bank promptly, exacerbating the well- The value of the payments made during known too-big-to-fail problem. Even if each of these costs is relatively small in normal times, their sum should be the day in a central bank’s large-value considered part of the tension generated by the link between payments system is typically far greater money and monetary policy. than the level of reserve balances held by banks overnight . . . . As a result, banks’ overnight reserve holdings are too small 3.2 Liquidity Policy to allow for the smooth functioning In times of stress or crisis in financial markets, the tension of the payments system during the day. between monetary policy and central banks’ other objectives can become acute. After the destructive events of September 11, 2001, the Federal Reserve recognized that the quantity of between a central bank’s monetary policy and its payments overnight reserves consistent with the target fed funds rate was policy. The central bank would like to increase the total supply too small to adequately address banks’ reluctance to make of reserve balances for payment purposes, but doing so would payments in a timely manner. The FOMC released a statement interfere with its monetary policy objectives. on September 17, 2001, that, in addition to lowering the target This tension has led to a common practice among central fed funds rate, stated: banks of supplying additional reserves to the banking system The Federal Reserve will continue to supply unusually for a limited time during the day. These daylight reserves (also large volumes of liquidity to the financial markets, called daylight credit) are typically lent directly to banks. Many as needed, until more normal market functioning is central banks provide daylight reserves against collateral at no restored. As a consequence, the FOMC recognizes that cost to banks. The Federal Reserve currently supplies daylight 10 the actual federal funds rate may be below its target credit to banks on an uncollateralized basis for a small fee. on occasion in these unusual circumstances.11 In providing daylight reserves, a central bank aims to allow banks to make their payments during the day smoothly and In this statement, the FOMC explicitly recognized the tension efficiently while limiting its own exposure to credit risk. between maintaining the market interest rate at its target level Under normal circumstances, this process of expanding and supplying more reserves to meet the demand for financial the supply of reserves during the day and shrinking it back market settlements. On September 18 and 19, the effective fed overnight works well; banks make payments smoothly and the funds rate was close to 1¼ percent while the target rate was central bank implements its target interest rate. However, this 3 percent. balancing act is not without costs. Lending large quantities of Exhibit 1 is again useful to help illustrate what happened. reserves to banks each day exposes the central bank to credit To meet the demand for reserves for financial settlements in risk. While requiring collateral for these loans mitigates credit various markets, the Fed increased the supply of reserve risk, it is an imperfect solution. If collateral is costly for banks balances. A shift in the supply curve to the right implies that to hold or create, the requirement imposes real costs. intersection with the demand curve will occur at a lower interest rate.12 In this case, it was not possible to achieve 10 See Board of Governors of the Federal Reserve System (2008b) for a proposal simultaneously the interest rate target and the increase in to change the Federal Reserve’s method of supplying daylight reserves. Under this proposal, banks would be able to obtain daylight reserves either on a overnight reserves necessary to ensure the efficient functioning collateralized basis at no cost or on an uncollateralized basis for a higher fee. For a general discussion of the Federal Reserve’s policies on daylight credit, 11 See
FRBNY Economic Policy Review / September 2008 47 of financial markets in conditions of stress. The exact same for the purpose of providing additional liquidity would lead to tension would arise under a symmetric channel system. Note, a lower short-term interest rate and, hence, would change the however, that the channel places a limit on how far the market stance of monetary policy. interest rate can deviate from the target—it cannot fall below Goodfriend (2002, p. 4) points out that central banks can the deposit rate. use other, less direct methods of managing broad liquidity: During the events of September 2001, the fed funds rate was below its target for only a few days and thus likely had no To some degree, the Fed can already manage broad liquidity under current operating procedures by changing impact on monetary policy objectives, as expectations were the composition of its assets, for example, by selling liquid short-term Treasury securities and acquiring less liquid The Federal Reserve faced a different longer term securities. However, the government debt injected into the economy in this way would not be as type of liquidity issue during the liquid as newly created base money. More importantly, financial market turmoil that began in the Fed’s ability to affect broad liquidity in this way is August 2007. In this case, there was strictly limited by the size of its balance sheet. a sharp decline in . . . broad liquidity: Interestingly, one of the new facilities introduced by the Fed in response to the market turmoil closely resembled the policy the ease with which assets in general described by Goodfriend. The Term Securities Lending can be sold or used as collateral at a price Facility, introduced in March 2008, provides loans of Treasury 13 that appropriately reflects the expected securities using less liquid assets as collateral. These loans increase broad liquidity by raising the total supply of highly value of the asset’s future dividends. liquid assets (reserves plus Treasury securities) in the hands of the private sector and decreasing the supply of less liquid assets. However, as Goodfriend observes, the amount of broad that the target rate would quickly be reestablished. It is an liquidity that can be provided through such a facility is strictly instructive episode, however, in that it demonstrates how limited by the quantity of Treasury securities owned by the increasing the supply of reserve balances available to the central bank. Thus, while a central bank can pursue a policy banking system can support market liquidity, and how this based on changes in the composition of its assets, such a policy objective can interfere with the maintenance of the target has inherent limitations. As we discuss in Section 4, alternative interest rate. methods of monetary policy implementation allow the central The Federal Reserve faced a different type of liquidity issue bank to overcome this limitation by pursuing a liquidity policy during the financial market turmoil that began in August 2007. based directly on bank reserves. In this case, there was a sharp decline in what Goodfriend (2002) calls broad liquidity: the ease with which assets in general can be sold or used as collateral at a price that appropriately reflects the expected value of the asset’s future dividends. 3.3 Efficient Allocation of Resources Goodfriend argues that increasing the supply of bank reserves can also support the level of broad liquidity in financial Another tension generated by the typical methods of monetary markets. This is especially true if the central bank uses the policy implementation described earlier relates to efficiency newly created reserves to purchase (or lend against) relatively concerns. These methods rely on banks facing an opportunity illiquid assets, thereby increasing the total quantity of liquid cost of holding reserves; their balances earn no interest in the assets held by the private sector. However, once again the link U.S. system and earn less than the prevailing market rate in a between money and monetary policy generates a tension; the symmetric corridor. This opportunity cost helps generate the central bank cannot pursue an independent “liquidity policy” downward-sloping part of the demand curve that the central using bank reserves. Any attempt to increase reserve balances bank uses to implement its target interest rate. The fact that
12 Needless to say, the disruption in financial markets would also tend to 13 The Fed also introduced other facilities, including the Term Auction Facility increase the demand for reserves, shifting the curve in Exhibit 1 to the right. and the Primary Dealer Credit Facility. Those facilities make loans of reserve The FOMC’s statement indicates a desire to more than compensate for this balances. In order to maintain the target interest rate, however, the Fed uses shift, that is, to increase reserve supply beyond the point that would maintain open market operations to “sterilize” these loans, leaving the total supply of the target interest rate given the increased reserve demand. reserve balances unaffected.
48 Divorcing Money from Monetary Policy Box 1 holding reserves is costly, however, conflicts with another Required Reserves central bank objective: the desire to promote the efficient Although this article emphasizes the similarities in monetary functioning of financial markets and the efficient allocation policy implementation procedures across countries, there are a of resources more generally. number of differences. One notable difference is in the use of Remunerating reserve balances at a below-market interest reserve requirements. Banks in the United States and the Euro zone rate is effectively a tax on holding these balances. (Box 1 are required to hold reserves in proportion to certain liabilities. In discusses how this tax is distortionary when applied to required reserves.) Similar logic shows that a distortion arises when other countries, including Australia and Canada, banks are not banks face an opportunity cost of holding excess reserves. required to hold any reserves; the only requirement is that a bank’s In this case, the tax leads banks to invest real resources in reserve account not be in overdraft at the end of the day. economizing on their holdings of excess reserves, but these In the simple framework we describe, it is immaterial whether efforts produce no social benefit. banks face a positive reserve requirement or the requirement is Reserve balances are costless for a central bank to create effectively zero. In reality, however, there are important differences through open market operations, for example, that exchange between these approaches. One such difference is that reserve newly created reserves for Treasury securities. If banks perceive requirements allow the central bank to implement reserve an opportunity cost of holding reserves (relative to Treasury averaging, whereby banks are allowed to meet their requirement on average over a reserve maintenance period rather than every day. As shown in Whitesell (2006a), reserve averaging tends to Another tension generated by the flatten the demand curve for reserves around the central bank’s target supply on all days of a maintenance period except the last typical methods of monetary policy one; this flattening tends to reduce volatility in the market interest implementation . . . relates to rate.a Another important difference is the extent of the distortions efficiency concerns. associated with bank reserve holdings. When required reserve balances do not earn interest, as is currently the case in the United States, the requirement acts as a tax on banks. This reserve tax raises securities, say), then they will engage in socially inefficient banks’ operating costs and drives a wedge between the price of efforts to reduce their use of reserves. In other words, the tax banking services and the social cost of producing those services, places a wedge between a private marginal rate of substitution creating a deadweight loss. The reserve tax also gives banks a strong and the corresponding social marginal rate of transformation. incentive to find ways to decrease their requirements, such as by This type of distortion was emphasized by Friedman (1959, pp. sweeping customers’ checking account balances on a daily basis 71-5), who argues that the central bank should pay interest on into other accounts not subject to reserve requirements. The all reserve balances at the prevailing market interest rate.14 efforts invested in these reserve-avoidance activities are clearly One might be tempted to suppose that the distortions wasted from a social point of view. created by this tax must be small because the quantity of excess Paying interest on required reserves at the prevailing market reserves held by banks is currently fairly small in the United rate of interest, as the European Central Bank does, eliminates States, around $1.5 billion. Such a conclusion is not warranted, most of these distortions. The Bank of England goes a step further however: the fact that the tax base is small does not imply that by having banks set voluntary balance targets. Once set, these the deadweight loss associated with the tax is insignificant. The targets can be used to implement monetary policy exactly the same deadweight loss includes all efforts banks expend to avoid way that reserve requirements are. However, because the targets holding excess reserves, including closely monitoring end-of- are chosen by the individual banks, rather than being determined day and end-of-maintenance-period balances so that any administratively, their creation generates none of the distortions associated with traditional reserve requirements. 14 This logic is central to the well-known Friedman rule, which calls for the central bank to eliminate the opportunity cost of holding all types of money a See Ennis and Keister (2008) for a detailed discussion of reserve (see especially Friedman [1969]). One way to implement this rule is by averaging in the type of framework used here. engineering a deflation that makes the real return on holding currency equal to the risk-free return. In this case, no interest needs to be paid on any form of money; the deflation generates the required positive return. In practice, there are a variety of concerns about deflation that keep central banks from following this approach. When applied to the narrower question of reserve balances held at the central bank, however, Friedman’s logic simply calls for remunerating all reserve balances at the risk-free rate.
FRBNY Economic Policy Review / September 2008 49 excess funds can be lent out, as well as actually lending the objectives. The tension created by these conflicts tends to be funds out. A substantial fraction of activity in the fed funds particularly strong during periods of stress in financial market is precisely of this type, and it is not clear whether markets. These tensions would be reduced or would disappear these indirect costs associated with the tax are small. The issues created by the reserve tax are sometimes described as a “hot potato” problem. Participants all try to get A central bank’s payments policy, liquidity rid of excess reserves because holding them is costly. However, the supply of excess reserve balances is fixed by the central bank policy, and desire to promote efficient and, at any point in time, someone must be holding them. allocation may all come into conflict Extending this analogy a bit, the fact that the potato itself (that with its monetary policy objectives. is, the quantity of excess reserve balances) is small does not imply that that the efforts spent passing it along are also small. The tension created by these conflicts This is especially true if the potato is very hot, that is, if excess tends to be particularly strong during reserve balances earn much less than the market rate of interest. periods of stress in financial markets.
The issues created by the reserve tax are altogether if banks did not face an opportunity cost of holding sometimes described as a “hot potato” overnight reserves that leads them to economize on their problem. Participants all try to get rid of holdings. In the next section, we describe an approach to implementing monetary policy that removes this opportunity excess reserves because holding them cost and discuss some of its implications. is costly. However, the supply of excess reserve balances is fixed by the central bank and, at any point in time, someone must be holding them. 4. Divorcing Money from Monetary Policy
Lucas (2000, p. 247) describes the deadweight loss associated The tensions we described all arise from the fact that, under with the inflation tax in a similar way: either current U.S. practice or a symmetric channel system, the quantity of reserve balances must be set to a particular level in In a monetary economy, it is in everyone’s private interest order for the central bank’s interest rate target to be achieved. to try to get someone else to hold non-interest-bearing There are, however, other approaches to monetary policy cash and reserves. But someone has to hold it all, so all implementation in which this strict link between money and of these efforts must simply cancel out. All of us spend monetary policy is not present. Here we discuss one such several hours per year in this effort, and we employ approach, which can be described as a floor-target channel thousands of talented and highly trained people to help system, or simply a floor system. This approach is a modified us. These person-hours are simply thrown away, wasted version of the channel system described above and has been on a task that should not have to be performed at all. advocated in various forms by Woodford (2000), Goodfriend (2002), Lacker (2006), and Whitesell (2006b). A particular type Any system of monetary policy implementation that relies of floor system has recently been adopted by the Reserve Bank on banks facing an opportunity cost of holding reserves of New Zealand. necessarily creates deadweight losses. The approaches described in the previous section thus conflict with a central bank’s desire to promote an efficient allocation of resources in the economy. 4.1 The Floor System We summarize by noting that a central bank’s payments policy, liquidity policy, and desire to promote efficient Starting from the symmetric channel system presented allocation may all come into conflict with its monetary policy in Exhibit 2, suppose that the central bank makes two
50 Divorcing Money from Monetary Policy Exhibit 3 overnight reserve balances are increased by the maximum A Floor System of Monetary Policy Implementation amount of current daylight credit use, then “in principle, any pattern of intraday payments that is feasible under the current Interest rate policy would still be feasible” even in the extreme case where access to daylight credit is eliminated altogether. Note that Lending rate restricting access to daylight credit will tend to increase the demand for overnight reserves, shifting the curve in Exhibit 3 Deposit rate Supply of to the right. The proposal in Lacker (2006) thus calls for equals reserves is target rate not linked to increasing the supply of reserves enough to ensure that it falls target rate on the flat portion of the demand curve even after this shift is Required or taken into account.16 reserves Goodfriend (2002) takes a different view, proposing that Reserve the supply of reserve balances could be used to stabilize 0 Target Target balances supply supply financial markets. The central bank could, for example, “increase bank reserves in response to a negative shock to broad liquidity in banking or securities markets or an increase in the external finance premium that elevated spreads in credit modifications. First, the deposit rate is set equal to the target markets” (p. 4). More generally, he suggests that the supply of rate, instead of below it. In other words, in this system the reserves could be set to provide the optimal quantity of broad central bank targets the floor of the channel, rather than some point in the interior. Second, the reserve supply is chosen so that it intersects the flat part of the demand curve generated A floor system “divorces” the quantity of by the deposit rate (Exhibit 3), rather than intersecting the money from the interest rate target and, downward-sloping part of the curve. Supply and demand will then cross exactly at the target rate, as desired.15 hence, from monetary policy. This divorce The key feature of this system is immediately apparent in gives the central bank two separate policy the exhibit: the equilibrium interest rate no longer depends on the exact quantity of reserve balances supplied. Any quantity instruments: the interest rate target can be that is large enough to fall on the flat portion of the demand set according to the usual monetary policy curve will implement the target rate. In this way, a floor system concerns, while the quantity of reserves “divorces” the quantity of money from the interest rate target and, hence, from monetary policy. This divorce gives the can be set independently. central bank two separate policy instruments: the interest rate target can be set according to the usual monetary policy concerns, while the quantity of reserves can be set liquidity services.17 It should be noted that there may be independently. complementarity between payments policy and liquidity policy If the quantity of reserves is no longer determined by with respect to reserve balances; increasing the reserve supply monetary policy concerns, how should it be set? In general, the to support broad liquidity can simultaneously reduce the use supply of overnight reserve balances could be used to ease any of daylight overdrafts, which might be particularly desirable of the tensions described earlier. For example, Lacker (2006) during times of market turmoil. suggests that increasing the supply of overnight reserves could The floor system also promotes a more efficient allocation reduce banks’ use of daylight credit without impairing their of resources. Not only does this approach eliminate the reserve ability to make timely payments. In fact, he argues that if tax, it also removes the opportunity cost of holding excess
16 15 The fact that these supply and demand curves cross at the target rate does not See Ennis and Weinberg (2007) for a formal analysis of the relationship imply that trades in the interbank market would occur at exactly this rate. A between daylight credit and monetary policy implementation, including the bank would require a small premium, reflecting transaction costs and perhaps ability of a floor system to reduce daylight credit usage. credit risk, in order to be willing to lend funds rather than simply hold them as 17 Determining this optimal quantity is a nontrivial task, however, and would (interest-bearing) reserves. As a result, the measured interest rate in the likely require more research on the notion of broad liquidity and its role in the interbank market would generally be slightly above the deposit rate. The target macroeconomy. The quantitative easing policy in place in Japan from 2001 to rate could instead be called the policy rate in order to make this distinction 2006 can be viewed as an attempt to use the supply of bank reserves to influence clear. macroeconomic outcomes.
FRBNY Economic Policy Review / September 2008 51 reserve balances. This is true for any quantity of reserve benefits—such as improved timeliness of payments—have balances large enough to lie on the flat portion of the demand already been observed. To be sure, the experience of a smaller curve in Exhibit 3. At such points, banks are indifferent at the country like New Zealand with this type of system may not be margin between reserves and other risk-free assets. As a result, directly applicable to other central banks. Nevertheless, it will they no longer have an incentive to invest real resources in be instructive to observe this experience and, in particular, to order to economize on their reserve holdings, and the see how it compares with the simple framework we present. deadweight loss associated with the systems described in Section 3 disappears. Woodford (2000) points to another advantage of the floor system. Suppose that innovation in financial markets were to 4.2 Discussion
While a floor system could potentially relieve or even eliminate the tensions between central bank objectives, there are several The Reserve Bank of New Zealand important concerns about how such a system would operate recently became the first central bank in practice and its potential effects on financial markets. One to implement a floor system. While it is concern is that a floor system would likely lead to a substantial reduction in activity in the overnight interbank market, as too early to evaluate the effects of this banks would have less need to target their reserve balance change properly, some benefits—such as precisely on a daily basis. In particular, since banks with excess improved timeliness of payments—have funds can earn the target rate by simply depositing them with the central bank, the incentive to lend these funds is lower than already been observed. it is under the other approaches to implementation discussed above. Nevertheless, an interbank market would still be necessary, as institutions will occasionally find themselves undermine the demand for reserve balances that is at the heart short of funds. How difficult it would be for institutions to of our model in Section 2. In particular, suppose that a perfect borrow at or near the target rate is an important open question. substitute for central bank reserves were developed and that In addition, some observers argue that the presence of an banks were able to avoid reserve requirements completely. In active overnight market generates valuable information and such a situation, the demand for reserves would fall to zero if there were any opportunity cost of holding them; banks would instead use the substitute private instrument for payment and While a floor system could potentially other liquidity purposes. If the central bank supplied a positive quantity of reserves, under the current system in the United relieve or even eliminate the tensions States the market interest rate would fall to zero. between central bank objectives, there Woodford argues that even in this extreme situation, the are several important concerns about central bank can still implement its target interest rate by using a floor system. Banks would again demand zero reserves at how such a system would operate any interest rate higher than the target rate in this situation. in practice and its potential effects However, under a floor system, the demand curve would be flat at the target rate for exactly the same reasons as described on financial markets. above. By setting a positive supply of reserves, therefore, the central bank could still drive the market interest rate to the target value. In this way, a floor system would enable the that some of this information would be lost if market activity central bank to meet its monetary policy objectives even if declined. For example, market participants must monitor the technological changes eliminated the special role currently creditworthiness of borrowers. If the overnight market were played by reserves; the key once again is divorcing money substantially less active, such monitoring may not take place on from monetary policy. a regular basis; this in turn could make borrowing even harder The Reserve Bank of New Zealand recently became the first for a bank that finds itself short of funds. Such monitoring may central bank to implement a floor system (Box 2). While it is also play a socially valuable role in exposing banks to market too early to evaluate the effects of this change properly, some discipline. It is important to bear in mind, however, that the
52 Divorcing Money from Monetary Policy Box 2 The Reserve Bank of New Zealand’s Floor System
In July 2006, the Reserve Bank of New Zealand (RBNZ) began collateral for Autorepo was reduced until the facility was the transition from a symmetric channel system of monetary policy discontinued on October 5. implementation to a floor system. We describe some reasons for Since the new framework was introduced, the RBNZ has the change and some features of the new regime, drawing heavily implemented two changes. First, banks are now allowed to use a wider on Nield (2006) and Nield and Groom (2008). set of assets to raise cash from the central bank. In particular, a limited From 1999 to 2006, the RBNZ operated a symmetric channel amount of AAA-rated paper is eligible.a Second, a tiered system of system with zero reserve requirements. It targeted a supply of NZD remuneration was introduced in response to episodes in which the 20 million overnight reserve balances every day. All reserve market interest rate rose substantially above the OCR. The RBNZ now balances were remunerated at a rate 25 basis points below the estimates the quantity of reserves a bank needs for its payment activity RBNZ’s target interest rate, called the official cash rate (OCR). and, based on this estimate, sets a limit on the quantity that will be Payments system participants could borrow reserves overnight remunerated at the OCR. Any reserves held in excess of that limit earn against collateral at the overnight reserve repurchase facility a rate 100 basis points below the OCR. This policy is designed to (ORRF), at a rate 25 basis points above the OCR. Finally, provide an incentive for banks to recirculate excess reserve positions participants could obtain reserves intraday, against collateral, and to prevent them from “hoarding” reserves. at an interest rate of zero using a facility called Autorepo. In principle, the RBNZ could have addressed this problem by The RBNZ’s decision to change the framework for monetary increasing its supply of reserves instead of by implementing a policy implementation followed signs of stress in the money tiered system. If the market interest rate is significantly higher than market. The Government of New Zealand had been running a the policy rate in a floor system, increasing the supply of reserves fiscal surplus for a number of years and government bonds had should drive the market rate down (see Exhibit 3 in the text). become increasingly scarce. The scarcity of government securities However, the RBNZ uses FX swaps to increase the supply of available to pledge in the Autorepo facility led to delayed payments reserves, and it found that the price in this market was moving between market participants. For the same reason, there had been against it; the more reserves the RBNZ created, the more costly an increase in the levels of underbid open market operations it became to create those reserves. It is worth noting that this and, consequently, in the use of the bank’s standing facilities at the problem would not arise in a country with a large supply of end of the day. Finally, the implied New Zealand dollar interest government bonds or with a central bank that can issue its own rates on overnight credit in the foreign exchange (FX) swap interest-bearing liabilities. In such cases, increasing the supply of market—the primary market by which banks in New Zealand reserves need not be costly and could be an attractive alternative traded overnight—were volatile and often significantly above to a tiered system. the target rate desired to implement monetary policy. While it is too early to evaluate with great confidence all of the The Reserve Bank of New Zealand conducted a review of its effects of the RBNZ’s changes, it appears that the transition went liquidity management regime in 2005 and announced the new smoothly overall. There were, of course, occasional signs of stress system in early 2006. Under this system, the RBNZ no longer in money markets, mostly attributable to the learning process offers daylight credit. In other words, there is no distinction experienced by the Bank and its payments system participants. between daylight and overnight reserves. The target supply of There are, however, definite positive signs that the liquidity of the reserves has been vastly increased to allow for the smooth interbank market has improved. Notably, payments have been operation of the payments system; the new level currently settling significantly earlier since the transition began, suggesting a fluctuates around NZD 8 billion. This represents an increase of reduction in the constraints previously attributable to the scarcity 400 times the level under the previous regime. Reserves are now of collateral available to pledge in the Autorepo facility. In remunerated at the OCR. It is still possible to obtain overnight addition, the implied New Zealand dollar interest rates in the FX funds at the ORRF, but at a rate 50 basis points above the OCR. swap market are now much less volatile and are well within the The bulk of the transition to this new system occurred in four 50 basis point band between the official cash rate and the ORRF. steps over a twelve-week period between July 3 and October 5, Finally, the RBNZ conducts open market operations much less 2006. During that time, the target supply of reserves increased frequently, and the operations are no longer subject to the gradually to its current level. At each step, the rate earned on underbidding that had led to excessive use of overnight facilities. reserves and the rate at which funds could be borrowed at the ORRF were increased relative to the OCR in increments of 5 basis a See the Reserve Bank of New Zealand’s May 2008 Financial Stability points up to their current levels. The set of securities eligible as Report for more details.
FRBNY Economic Policy Review / September 2008 53 market for overnight loans of reserves differs from other develop and how efficiently they would operate are important markets in fundamental ways. As we discussed, reserves are not open questions. a commodity that is physically scarce; they can be costlessly Going forward, the experience of New Zealand’s floor produced by the central bank from other risk-free assets. system will provide valuable information on these issues and Moreover, there is no role for socially useful price discovery others that might arise. However, the differences between the in this market, because the central bank’s objective is to set a financial system of New Zealand and those of economies particular price. Weighing the costs and benefits of a reduction like the United States will make it difficult to draw definite in market activity is therefore a nontrivial task and an conclusions. For this reason, it is important to employ the important area for future research. tools of modern economic theory to develop models that are If desired, the floor system could be modified in ways that capable of addressing these issues. encourage higher levels of activity in the overnight interbank market. For example, the central bank could limit the quantity of reserves on which each bank earns the target rate of interest and compensate balances above this limit at a lower rate. Such limits would encourage banks that accumulate unusually large 5.Conclusion balances over the course of the day to lend them out. By setting lower limits, the central bank would encourage more activity This article highlights the important similarities in the in the interbank market while marginally increasing the monetary policy implementation systems used by many central distortions discussed above.18 Whitesell (2006b) presents a banks. In these systems, there is a tight link between money and monetary policy because the supply of reserve balances must be set precisely in order to implement the target interest rate. This If desired, the floor system could be link creates tensions with the central bank’s other objectives. modified in ways that encourage higher For example, the intraday need for reserves for payment levels of activity in the overnight purposes is much higher than the overnight demand, which has led central banks to provide low-cost intraday loans of reserves interbank market. to participants in their payments systems. This activity exposes the central bank to credit risk and may generate problems of moral hazard. The link also prevents central banks from system in which banks are allowed to determine their own increasing the supply of reserves to promote market liquidity in limits by paying a “capacity fee” proportional to the chosen times of financial stress without compromising their monetary limit. In this case, the central bank would set the fee schedule in policy objectives. Furthermore, the link relies on banks facing a way that balances concerns about the level of market activity an opportunity cost of holding reserves, which generates with the resulting level of distortions. deadweight losses and hinders the efficient allocation of Another interesting issue is the extent to which a floor resources. system would allow the central bank to restrict access to Our study also presents an approach to implementing daylight credit, if it so desired. If access to daylight credit is monetary policy in which this link is severed, leaving the substantially restricted or removed, the smooth functioning of quantity of reserves and the interest rate target to be set the payments system may require banks to acquire funds in the independently. In this floor-system approach, interest is paid market on a timely basis during the day. In principle, this could on reserve balances at the target interest rate. This policy allows be accomplished by the development of either an intraday the central bank to increase the supply of reserves, perhaps even market for reserve balances or a market for precise time-of-day significantly, without affecting the short-term interest rate. delivery of reserves (see McAndrews [2006] for a discussion While the floor system has received a fair amount of attention of such possibilities). Whether such markets would actually in policy circles recently, there are important open questions 18 Ennis and Keister (2008) describe a related approach based on “clearing about how well such a system will work in practice. Going bands,” where banks face a minimum requirement and earn the target rate of forward, it will be useful to develop theoretical models of the interest on balances held up to a higher limit. This approach could be used to monetary policy implementation process that can address encourage activity in the interbank market on the borrowing side (by banks that find themselves below the minimum requirement) as well as on the these questions, as well as to observe New Zealand’s experience lending side (by banks that find themselves above the higher limit). with the floor system it implemented in 2006.
54 Divorcing Money from Monetary Policy References
Bartolini, L., G. Bertola, and A. Prati. 2002. “Day-to-Day Monetary Friedman, M. 1959. A Program for Monetary Stability. Policy and the Volatility of the Federal Funds Interest Rate.” New York: Fordham University Press. Journal of Money, Credit, and Banking 34, no. 1 (February): 137-59. ———. 1969. The Optimum Quantity of Money and Other Essays. Chicago: Aldine Publishing Company. Bech, M. L., and B. Hobijn. 2007. “Technology Diffusion within Central Banking: The Case of Real-Time Gross Settlement.” Furfine, C. H. 2000. “Interbank Payments and the Daily Federal Funds International Journal of Central Banking 3, no. 3 Rate.” Journal of Monetary Economics 46, no. 2 (October): (September): 147-81. 535-53.
Board of Governors of the Federal Reserve System. 2007. “Overview Goodfriend, M. 2002. “Interest on Reserves and Monetary Policy.” of the Federal Reserve’s Payments System Risk Policy on Federal Reserve Bank of New York Economic Policy Review 8, Daylight Credit.” Sixth Edition. July. Available at
———. 2008a. “Monetary Policy Report to the Congress.” Guthrie, G., and J. Wright. 2000. “Open Mouth Operations.” Journal February 27. Available at
Ennis, H. M., and T. Keister. 2008. “Understanding Monetary Policy McAndrews, J. J., and S. M. Potter. 2002. “Liquidity Effects of the Implementation.” Federal Reserve Bank of Richmond Economic Events of September 11, 2001.” Federal Reserve Bank of New York Quarterly 94, no. 3 (summer). Economic Policy Review 8, no. 2 (November): 59-79.
Ennis, H. M., and J. A. Weinberg. 2007. “Interest on Reserves and Nield, I. 2006. “Changes to the Liquidity Management Regime.” Daylight Credit.” Federal Reserve Bank of Richmond Economic Reserve Bank of New Zealand Bulletin 69, no. 4 (December): Quarterly 93, no. 2 (spring): 111-42. 26-31.
FRBNY Economic Policy Review / September 2008 55 References (Continued)
Nield, I., and J. Groom. 2008. “Implementing Monetary Policy in New ———. 2006b. “Monetary Policy Implementation without Averaging Zealand.” Unpublished paper, Reserve Bank of New Zealand. or Rate Corridors.” Board of Governors of the Federal Reserve System Finance and Economics Discussion Series, Poole, W. 1968. “Commercial Bank Reserve Management in a no. 2006-22, May. Stochastic Model: Implications for Monetary Policy.” Journal of Finance 23, no. 5 (December): 769-91. Woodford, M. 2000. “Monetary Policy in a World without Money.” International Finance 3, no. 2 (July): 229-60. Thornton, D. L. 2006. “The Daily Liquidity Effect.” Federal Reserve Bank of St. Louis Working Paper no. 2006-020A, April.
Whitesell, W. 2006a. “Interest Rate Corridors and Reserves.” Journal of Monetary Economics 53, no. 6 (September): 1177-95.
56 Divorcing Money from Monetary Policy Empirical Analyses of Trends in Large-Value Payments
Articles by:
Morten L. Bech, Christine Preisig, and Kimmo Soramäki Olivier Armantier, Jeffrey Arnold, and James McAndrews Christopher Becher, Marco Galbiati, and Merxe Tudela
Morten L. Bech, Christine Preisig, and Kimmo Soramäki
Global Trends in Large-Value Payments
• The evolving landscape in which large-value 1.Introduction payments systems (LVPSs) operate is having important effects on the financial system. lobalization and technological innovation are two of the G most pervasive forces affecting the financial system and • An analysis of the current interbank payment its infrastructure. Perhaps nowhere are these trends more environment points to three forces that are apparent than in the internationalization and automation of shaping ten trends common to LVPSs payments. The evolving landscape is most obvious in retail around the world. payments. The use of paper checks is in rapid decline or has been eliminated in most of the industrialized world. Credit and • Technological innovation is making LVPSs debit cards can be used in the most surprising places. Internet safer and more efficient while allowing for banking with money transfer capabilities is common, and new systems that are not limited to one several providers are competing to service consumers’ payments over the Internet and mobile devices. country or currency. In wholesale, or interbank, payments, the effect of globalization and technological innovation is probably less Structural changes in banking—such as • obvious to the casual observer—but it has been equally immense growth in the financial sector, impressive. Given the importance of payments and settlement changes in the role of firms and their products, systems to the smooth operation as well as resiliency of the and greater globalization of financial financial system, stakeholders need to understand and assess institutions and their services—are influencing the potential consequences of this evolution. This article offers the use of LVPSs. an in-depth look at the current environment for large-value payments systems (LVPSs). We describe ten trends common to • The evolution of central bank policies is LVPSs around the world and identify the key drivers of these resulting in central banks becoming more developments and the most important policy issues facing active in monitoring existing and planned central banks (see box). Furthermore, we provide empirical systems, assessing systems according to support for each of the trends by using numerous publicly international standards, and inducing change. available sources, including Bank for International Settlements
Morten L. Bech is a senior economist at the Federal Reserve Bank of New York; The views expressed are those of the authors and do not necessarily reflect the Christine Preisig is a former senior policy expert at the Bank; Kimmo Soramäki position of the European Central Bank, the Federal Reserve Bank of New York, is a Ph.D. candidate at Helsinki University of Technology on leave from the or the Federal Reserve System. Any errors or omissions are the responsibility European Central Bank. This article was written while Mr. Soramäki was on of the authors. leave from the European Central Bank and acting as a research consultant to the Federal Reserve Bank of New York. Correspondence:
7. Falling Numbers of System Participants have motivated them to develop hybrid systems. Developments 3. Evolution of in payments system technology have also facilitated the Central Bank 8. Extended Operating Hours Policies emergence of systems that settle payments across national 9. Declining Transaction Fees borders in one or more currencies. In addition, the clearing of 10. Adoption of Common Standards payments is in some instances moving offshore and the ability for Large-Value Payments Systems of participants to connect remotely—eliminating the need for a physical “footprint” in the jurisdiction of LVPSs—is becoming more widespread. Foreign exchange (FX) settlement and counterparty risk are being managed more tightly in (BIS) statistics on payments and settlement systems in selected part because of the use of payment-versus-payment (PvP) countries (the “Red Book”). We focus on large-value payments systems in countries where the central bank is a member of Technological innovation, structural the Committee on Payment and Settlement Systems (CPSS), a body under the auspices of the BIS (Appendix A). changes in banking, and the evolution Technological innovation, structural changes in banking, of central bank policies are the three main and the evolution of central bank policies are the three main reasons for the recent developments in large-value payments. reasons for the recent developments First, technological innovation has created opportunities to in large-value payments. make existing large-value payments systems safer and more efficient. Such innovation has also accommodated the 1 industry’s growing need for new types of systems that are not mechanisms. CLS Bank operates a multicurrency payments limited to a single country or a currency. Second, the financial system for the simultaneous settlement of both sides of a sector has experienced immense growth over the last few foreign exchange transaction on a PvP basis. With CLS Bank, decades accompanied by changes in the role of individual firms existing risks associated with FX trades are virtually eliminated. and the products they offer. In addition, financial institutions The next three trends—increasing settlement values and and their services have become increasingly globalized. These volumes, shrinking average payment sizes, and falling numbers structural changes have affected how participants use large- of system participants—as well as the emergence of cross- value payments systems. Third, the role of central banks in border and offshore systems (Trend 3) fall into the second large-value payments systems has changed significantly in category. They are determined largely by how the banking recent years. Central banks have become more involved in sector uses payments systems and by the structural changes payments systems and have created formal and systematic taking place therein. The values and volumes originated over oversight functions. The main focus lies in promoting safety LVPSs grew exponentially until the turn of the century. and efficiency in LVPSs and in maintaining overall financial However, in terms of value, growth has since slowed and is no stability. Central banks therefore have taken more active roles longer outpacing economic growth as measured by GDP. in monitoring existing and planned systems, in assessing 1 PvP ensures that a final transfer of one currency occurs only if a final transfer systems according to international standards, and, if necessary, of the other currency or currencies takes place (Bank for International in inducing change. Settlements 2003).
60 Global Trends in Large-Value Payments Because many LVPSs process a large amount of relatively Over the last few decades, many countries have chosen to low-value payments, the average payment size settled has modify the settlement procedure employed by their interbank shrunk. Hence, the dichotomy between small- and large-value payments system with a view to reducing settlement risks and payments systems is not always applicable. In addition, the potential for adverse systemwide implications. Most central consolidation in the banking sector has led to fewer banks have opted for the implementation of an RTGS system. participants in LVPSs. Structural changes have also resulted in Such a system reduces settlement risk, as payments are settled the emergence of global banks that require a global payment individually and irrevocably on a gross basis in real time, infrastructure, which in turn has led to the creation of new ensuring immediate finality. RTGS can also help reduce systems that accommodate these needs. settlement risk by facilitating payment versus payment and The last three trends and the rise of CLS Bank (Trend 4) fall delivery versus payment in the settlement of FX and securities into the third category. They are associated with central banks’ transactions, respectively. operating policies regarding LVPSs. The service level of all Fedwire is the world’s oldest RTGS system. Its origins can systems is improving with longer operating hours. Some be traced to 1918, when the Federal Reserve inaugurated a systems are even approaching a twenty-four-hour settlement network of wire communications among the individual cycle. Transaction costs in various LVPSs have been falling since the late 1990s because the savings achieved through improvements in operating efficiency have been passed on to Over the last few decades, many countries system participants in the form of lower fees. Through the have chosen to modify the settlement adoption of common standards, such as the CPSS’ Core Principles for Systemically Important Payments Systems, risk procedure employed by their interbank management in LVPSs has become more standardized. payments system with a view to reducing Furthermore, the central bank community was the driving force behind the development of CLS Bank. settlement risks and the potential for We now describe each trend in detail and conclude by adverse systemwide implications. commenting on the possible future of the large-value payments Most central banks have opted for the landscape. implementation of an RTGS [real-time gross settlement] system. 2.Trend 1: Diffusion of Real-Time Gross Settlement Systems2 Reserve Banks. In the early 1970s, the Fedwire system migrated to a fully computerized platform, and settlement in “real time” As a consequence of the rapid increase in values settled in large- was achieved. value payments systems in the 1980s, central banks became A number of western European countries began concerned about settlement risks inherent in the then- implementing RTGS systems in the 1980s. By 1988, RTGS 3 prevalent deferred net settlement systems. In particular, the systems operated in four of the six major currencies. RTGS banks were concerned about the potential for contagion (or adoption continued at a rate of roughly one country per year even a systemic event) attributable to the unwinding of the net during the early 1990s. In 1992, the Treaty of Maastricht positions that would result if a participant failed to make good created the foundation for the Economic and Monetary Union 4 on its obligations when due. (EMU). A year later, the central banks in the European Union (EU) agreed that each member state should have an RTGS 2 For more on these systems, see Bech (2007) and Bech and Hobijn (2007). system. Furthermore, in 1995 it was decided to interlink the 3 A DNS system effects the settlement of obligations or transfers between or among counterparties on a net basis at some later time (Bank for International national RTGS systems through the Trans-European Settlements 2003). Automated Real-time Gross settlement Express Transfer 4 Unwinding is a procedure followed in certain clearing and settlement systems (TARGET) system to facilitate the European Central Bank’s in which transfers of securities and funds are settled on a net basis, at the end (ECB) single monetary policy and to promote sound and of the processing cycle, with all transfers provisional until all participants have discharged their settlement obligations. If a participant fails to settle, some or efficient payment mechanisms in euros. This decision led to a all of the provisional transfers involving that participant are deleted from the flurry of new systems and upgrades to existing ones. TARGET system and the settlement obligations from the remaining transfers are then went live on January 4, 1999, and even EU countries that did recalculated. Such a procedure has the effect of allocating liquidity pressures and losses attributable to the failure to settle to the counterparties of the not join the EMU at the outset (the United Kingdom, participant that fails to settle (Bank for International Settlements 2003). Denmark, and Sweden) were allowed to participate in the
FRBNY Economic Policy Review / September 2008 61 system. As the ECB made RTGS a prerequisite for membership 174 central banks were using RTGS systems. The RTGS in the EMU, prospective members in the rest of Europe began adoption rate was about one central bank per year in the latter to implement RTGS. Furthermore, as hostilities ended in the part of the 1980s and the beginning of the 1990s. In the Balkans in the late 1990s, governments began to rebuild their mid-1990s, the rate accelerated with the addition of three respective economies. They considered the establishment of central banks in 1995 and five in 1996. Since then, the annual sound and efficient financial systems a priority. RTGS systems adoption rate has not dipped below three new central banks. were implemented with support from the EU, the International It peaked in 2002, when a total of fifteen central banks Monetary Fund (IMF), and the World Bank. With ongoing implemented new RTGS systems. projects in Russia and Cyprus, the diffusion of RTGS in Europe is nearly completed. Outside Europe, the rate of RTGS adoption since the mid- 1990s has been equally impressive. Australia and New Zealand 3.Trend 2: Take-off of Hybrid implemented RTGS in 1998. In Asia, the rate of implementa- Systems tion has been fairly steady; on average, about one country per year has adopted RTGS. Six countries in the Middle East have As payments are settled individually in an RTGS system, done likewise. In Africa, the South African Reserve Bank sufficient liquidity needs to be available to fund each payment. (SARB) spearheaded RTGS adoption in 1998. Through the Real-time gross settlement thus reduces settlement risks but South African Development Community (SADC),5 SARB has results in an increased need for intraday liquidity to smooth 7 participated in developing and strengthening the financial nonsynchronized payment flows. infrastructure in the rest of southern Africa. As of 2006, eleven Initially, central banks provided intraday credit free to African central banks have implemented RTGS, many with the commercial banks. This policy is no longer considered a viable support of the World Bank. option by the banks, as it exposes them (and ultimately In the Western Hemisphere, Canada is the only Group of Ten (G-10) country that has decided not to implement an RTGS system. Instead, Canada opted for a hybrid system.6 To reduce the need for intraday liquidity, Uruguay was the first country in South America to adopt several systems have developed different RTGS in 1995. By 2006, seven of thirteen South American types of queue management and countries had followed suit. Implementation in Central America and the Caribbean has started only recently, but the liquidity-saving features. Inter-American Development Bank is assisting RTGS efforts in the region. taxpayers), as guarantor of the finality of payments, to credit The global diffusion of RTGS systems since the mid-1980s risk (see, for example, Humphrey [1986] and Bech and is evident from Exhibit 1. By 1985, three central banks— Soramäki [2005]). Thus, intraday liquidity is costly for the Federal Reserve, Danmarks Nationalbank, and the participants either in the form of explicit fees or implicitly as Netherlandsche Bank—had implemented RTGS systems. the opportunity cost of collateral that participants need to A decade later, that number had increased to sixteen, but RTGS pledge for an intraday credit line from the central bank. To was still utilized predominantly by industrialized countries. reduce the need for intraday liquidity, several systems have In recent years, however, transitional as well as developing developed different types of queue management and liquidity- countries have begun investing heavily in improving their saving features. The queue management features include financial systems, and now RTGS is a common choice for different priority categories for payments and the possibility interbank payments. At the end of 2006, 93 of the world’s of reordering payments once in the queue. The liquidity-saving
5 The member states of the SADC are Angola, Botswana, Democratic Republic 7 Payments systems can operate at different levels of liquidity requirements, of Congo, Lesotho, Malawi, Mauritius, Mozambique, Namibia, Seychelles, delays, and risks. Depending on the system design, these can be traded off South Africa, Swaziland, Tanzania, Zambia. and Zimbabwe. against each other. Liquidity requirements are highest when payments are 6 The Canadian Large-Value Transfer System (LVTS) processes payments with settled continuously against full cover (as in RTGS). Liquidity requirements in finality in real time, while settlement occurs on a multilateral net basis at the an RTGS system can be reduced by delaying payments until incoming end of the day. Immediate intraday finality is achieved because settlement is payments allow settlement. In deferred net settlement systems, payments are guaranteed under all circumstances. This is facilitated by the use of collateral to settled only periodically by transferring only net amounts, thus payments are secure participants’ intraday net debit (negative) positions and by a residual delayed from time of receipt until time of settlement. In such systems, banks guarantee provided by the Bank of Canada (see Arjani and McVanel [2006]). can reduce delays by crediting customer accounts before final settlement. This LVTS is considered equivalent to RTGS in terms of finality, as the Bank of will, however, come at the expense of credit risks, as final settlement may not Canada provides an explicit guarantee of settlement in case of participant take place as expected. For an in-depth discussion of these trade-offs, see failure. Leinonen and Soramäki (1999).
62 Global Trends in Large-Value Payments Exhibit 1 Diffusion of Real-Time Gross Settlement Systems Worldwide
1985
1995
2006
Source: Bech and Hobijn (2007).
FRBNY Economic Policy Review / September 2008 63 Chart 1 Payments Settled in RTGS, DNS, and Hybrid Modes, 1999 and 2005 U.S. Dollar Equivalent
1999 Hybrid 2005 DNS $41 trillion $53 trillion (3 percent) (3 percent)
Hybrid DNS $612 trillion $645 trillion (32 percent) RTGS (46 percent) $727 trillion (51 percent) RTGS $1,274 trillion (65 percent)
Sources: Bank for International Settlements; authors’ calculations. Notes: RTGS is real-time gross settlement; DNS is deferred net settlement. Figures represent Fedwire, CHIPS, CHAPS, SIC, TARGET, MEPS, K-RIX, LVTS, PNS, Euro1, and HKD CHATS. DNS in 1999: Euro1 and BoJ-NET. Hybrids in 2005: PNS, LVTS, CHIPS, and RTGSplus. Systems are described in Appendix A. The value for RTGSplus was subtracted from the value for TARGET in computing the value settled via RTGS.
features typically involve the netting of payments in the queue, employ advanced bilateral or multilateral offsetting algorithms a feature commonly referred to as gridlock resolution (see Bech on a continuous basis. Prominent examples of hybrid systems and Soramäki [2001]).8 include RTGSplus in Germany, LVTS in Canada, CHIPS in the Another approach to balancing risks, payment delays, and United States, and Paris Net Settlement (PNS) in France. liquidity needs more efficiently has been the development of With regard to settlement method, the major development hybrid systems (see McAndrews and Trundle [2001] and in payments systems in CPSS countries has been the sharp Leinonen and Soramäki [1999]). Hybrid systems employ increase in value settled by hybrid systems (Chart 1). In 1999, advanced settlement algorithms that combine components of 3 percent of payment value was settled by these systems. RTGS both net and real-time gross settlement. Some payments may settlement accounted for approximately 50 percent of be settled individually, as in RTGS, while others, usually less payments and DNS settlement for roughly 45 percent.10 urgent payments, may be pooled together and netted.9 Other By 2005, hybrid systems accounted for close to one-third of features may include bilateral limits to manage credit value settled, whereas RTGS increased to almost two-thirds. exposures and reciprocity. The distinction between RTGS The only remaining DNS system, EURO1, accounted for less systems and hybrid systems can be fluid. In this article, hybrid than 3 percent of total value settled in CPSS large-value systems are defined as systems with either separate payment payments systems. streams for urgent or nonurgent payments and/or systems that
8 Gridlock is a situation that can arise in a funds or securities transfer system when the failure of certain transfer instructions to be executed (because the needed funds or securities balances are unavailable) prevents the execution 10 Prior to 2001, BoJ-NET provided both DNS and RTGS settlement modes. of a substantial number of instructions from other participants (Bank for However, the RTGS settlement mode was seldom used by banks because of its International Settlements 2003). higher liquidity costs. In 2001, the Bank of Japan introduced a reconfigured 9 For example, the RTGS share of the German hybrid system RTGSplus is about BoJ-NET and abolished DNS. Just before the change, only 3 percent of Japan’s 10 percent, whereas the netted payments share is 90 percent. wholesale payments were settled via RTGS (Selgin 2004).
64 Global Trends in Large-Value Payments 4.Trend 3: Emergence of Fedwire and CHIPS in the United States, CHAPS in the Cross-Border and Offshore United Kingdom, and BoJ-NET in Japan, to name a few, are Payments Systems traditional large-value payments systems that settle payments in the local currency for domestic participants. Large-value payments systems have traditionally settled Originally, participation in the Swiss Interbank Clearing payments in the local currency among participants located (SIC) system was limited to banks domiciled in Switzerland within the same national borders as the system. However, since and the Principality of Liechtenstein. However, since 1998 the late 1990s, systems have emerged that allow payments to remote access to SIC has been granted to banks domiciled settle across national borders, facilitate settlement in multiple outside Switzerland. Any bank worldwide can participate in currencies, and permit participants to be located in a foreign SIC as long as it meets the admission criteria (see Heller, jurisdiction. The two key drivers of the development of cross- Nellen, and Sturm [2000]). Among other things, remote access border and offshore systems have been the introduction of the has allowed foreign banks that participate in the futures and euro and the demand for payment settlement in foreign currencies as part of payment-versus-payment FX transactions, particularly in the Asian time zones. The two key drivers of the development of To analyze these new types of systems, we classify them according to the location of their participants and the currencies cross-border and offshore systems have in which they settle payments. A simple taxonomy is presented been the introduction of the euro and the in Table 1. The type of participants and currencies that a system services are each divided into three groups. Participants are demand for payment settlement in foreign categorized as domestic, remote, or cross-border, whereas the currencies as part of payment-versus- currencies settled are classified as local, foreign, or multiple. payment FX transactions, particularly We refer to participants located in the same country as the system as domestic participants. A remote participant has in the Asian time zones. neither its head office nor any of its branches located in the country where the transfer system is based (Bank for International Settlements 2003). Cross-border participants are options exchange EUREX to process Swiss franc transactions payees and payers located in different countries. The groups are directly via SIC without having a physical “footprint” in not exclusive, as systems with remote or cross-border Switzerland. At the end of 2006, of the 331 SIC participants, participants also service domestic participants. Further on, we 72 were so-called remote members. distinguish between systems that settle in the local currency of In 1999, with the introduction of the euro, two pan- the country where the system is located, systems that settle in a European interbank payments systems were introduced. single foreign currency, and systems that settle in a set of TARGET and EURO1 both settle cross-border payments in multiple currencies that typically includes the local currency. euros. Currently, TARGET consists of 17 RTGS systems with 1,058 direct participants11 (see Appendix B for more on TARGET and its individual components). In 2006, TARGET Table 1 processed on average 326,000 daily payments worth about Taxonomy of Payments Systems 2.1 trillion euros (see Appendix C for more on cross-border payments within TARGET). Settlement Currency In conjunction with EU central banks’ efforts, the European Multiple (PvP in Foreign Exchange Banking Association (EBA) established, much like CHIPS, a Participant Type Local Foreign Settlement) private sector complement to TARGET: the EURO1 system.
Domestic Fedwire, CHIPS, USD CHATS, HKD, USD, Today, the system has 70 participating banks and processes on CHAPS Sterling, EUR CHATS and EUR average 185,000 payments a day with a total value of around LVTS, RIX, PNS, CHATS (PvP 195 billion euros. TARGET and EURO1 are examples of LVPSs BoJ-NET arrangement in CHATS) that settle cross-border payments in a local currency.
Remote SIC Euro-SIC — 11 Cross-border TARGET, CHAPS Euro CLS Bank, The systems are Belgium ELLIPS (Belgium), KRONOS (Denmark), EURO1 RENTAS-CHATS RTGSplus (Germany), HERMES (Greece), SLBE (Spain), EP RTGS (Estonia), link TBF (France), IRIS (Ireland), New BIREL (Italy), LIPS-Gross (Luxembourg), TOP (the Netherlands), Artis (Austria), SORBNET-EURO (Poland), SPGT Notes: Systems are described in Appendix A. PvP is payment-versus- (Portugal), Payments System (Slovenia), BoF-RTGS (Finland), and CHAPS payment. (United Kingdom).
FRBNY Economic Policy Review / September 2008 65 The United Kingdom, Sweden, and Denmark did not join Chart 2 the EMU at the outset. However, in expectation that these Annual Value Settled by U.S. Dollar countries would eventually join, separate euro RTGS systems Offshore Systems, 1999-2006 were built and connected to TARGET: CHAPS Euro in the Billions of U.S. dollars United Kingdom, E-RIX in Sweden, and DEBES in Denmark. 2,000 From the perspective of the Table 1 taxonomy, these systems USD CHATS can be thought of as facilitating cross-border payments in a 1,500 foreign currency. However, with the advent of the next BM&F generation of the TARGET system, the euro functionality of all 1,000 three systems will be phased out. E-RIX already closed at the end of 2006. INDIA USD Another implication of the common currency was the 500 introduction of euroSIC in 1999. The system allows Swiss banks to conduct euro transactions. It operates on the same 0 1999 00 01 02 03 04 05 06 platform as the Swiss franc system, but settlement takes place on the books of the Swiss Euro Clearing Bank (SECB) in Sources: Hong Kong Monetary Authority; Clearing Corporation Frankfurt, Germany. SECB provides a link to the euro area by of India, Limited; Bolsa de Mercadorias & Futuros. being a direct participant in RTGSplus, through which access to TARGET is established.12 In addition, remote access is possible in euroSIC. Within the taxonomy, euroSIC is a system in the foreign currency with correspondent banks that settles a foreign currency for domestic and remote abroad. The transfer of funds takes place within the participants. same settlement window. CLS Bank operates a system that settles multiple currencies • In April 2003, HKICL introduced EUR CHATS for the for participants located in different countries (we discuss it in settlement of euro payments. The system is similar to more detail as part of Trend 4). A similar system is the link USD CHATS, with the settlement bank being Standard between the ringgit real-time gross settlement system in Chartered in London. A novel feature of the combined Malaysia (the RENTAS system) and the U.S. dollar real-time CHATS systems is that it allows for PvP settlement of gross settlement system in Hong Kong (USD CHATS). The USD-HKD and EUR-HKD as well as USD-EUR foreign exchange trades. In the Table 1 taxonomy, this system link, established by the Hong Kong Monetary Authority and enables the settlement of multiple currencies for Bank Negara Malaysia in 2006, allows for PvP foreign exchange domestic system participants. settlement of USD-MYR trades. In the Table 1 taxonomy, the link can be considered a cross-border PvP system. • In August 2003, Clearing Corporation of India, Ltd. (CCIL), introduced a system that clears and settles Since 2000, several new systems that settle a foreign interbank FX trades, including Indian rupee (INR) and currency for participants located within the same national USD. CCIL is a third-party member of CLS Bank and borders as the system itself have emerged: currently uses ABN AMRO in New York as its settlement • In 2000, Hong Kong Interbank Clearing Ltd. (HKICL) bank.13 introduced the U.S. Dollar Clearing House Automated Many of these systems can be understood as substitutes for Transfer System (USD CHATS) to clear USD payments traditional correspondent banking services. The systems offer a in the Asian time zone. The system is technically the customized local service in the native language, operating same as the Hong Kong dollar CHATS system, except hours that accommodate their customers’ needs, and that settlement currently takes place on the books of a private bank (HSBC) in New York City. potentially better risk management through the use of more formal rules and procedures. • In April 2002, the Brazilian Bolsa de Mercadorias & We refer to systems that settle a foreign currency for Futuros (BM&F) FX Clearinghouse initiated operations domestic, remote, or cross-border participants as offshore for settling USD-BRL trades. The Clearinghouse systems. Offshore systems settling USD have experienced maintains a settlement account in the local currency with the Central Bank of Brazil and settlement accounts strong growth rates, albeit from a low initial level (Chart 2). All three USD offshore systems settled about USD 1,500 billion 12 Interestingly, the largest Swiss bank, UBS AG, processes its cross-border euro payments via proprietary access to RTGSplus and not via euroSIC (see the 13 Other offshore USD systems in operation are the Moscow Interbank interview with Stephan Zimmerman, Head of Operations at UBS AG, at Currency Exchange in Russia and the Philippine Domestic Dollar Transfer
66 Global Trends in Large-Value Payments Chart 3 take measures to control their FX settlement exposures by Annual Value Settled by Euro Offshore Systems, improving their practices for measuring and managing 1999-2006 exposures. Second, industry groups were encouraged to
Billions of euros Billions of euros develop well-constructed multicurrency services that would 3,000 40,000 contribute to the risk-reduction efforts of individual banks. 35,000 2,500 CHAPS euro Third, central banks committed themselves to encouraging and Scale 30,000 fostering private sector development in this field. They also 2,000 25,000 agreed to improve national payments systems to facilitate E-RIX private sector risk-reduction efforts. 1,500 Scale 20,000 Since 1996, there has been considerable progress on all three 15,000 1,000 euroSIC tracks of the strategy. In particular, the launch of CLS Bank in Scale 10,000 September 2002 was a significant move forward by the industry 500 5,000 EUR CHATS to reduce foreign exchange settlement risk. CLS Bank operates Scale 0 0 a multicurrency system that settles payments for participants 1999 00 01 02 03 04 05 06 located on every continent. As transactions are settled on a PvP Sources: Bank for International Settlements; authors’ calculations. basis, Herstatt risk associated with these trades is virtually eliminated. CLS Bank is a special-purpose U.S. bank supervised by the Federal Reserve and under the cooperative oversight of in 2006. In euro offshore systems, growth has been more the central banks of the fifteen currencies included in the modest—especially considering the fact that the fastest- system. It has grown extremely fast and is on par with Fedwire growing systems, CHAPS Euro and E-RIX, have been and TARGET in terms of value settled (Chart 4). discontinued (Chart 3). All of the offshore systems (excluding On March 19, 2008, CLS Bank settled a record 1,113,464 CHAPS Euro), however, are still of smaller orders of magni- payment instructions with a gross value of USD 10.3 trillion. tude than the smallest domestic LVPS in the CPSS countries. CLS Bank has grown steadily since its inception as a result of increasing FX volumes, new currencies, and greater market penetration. It now settles more than USD 200 trillion per quarter and had surpassed both Fedwire and TARGET in terms 5.Trend 4: The Rise of Continuous of value settled in the second half of 2005. Chart 4 also displays Linked Settlement Bank the euro and USD values settled in CLS Bank. Whereas TARGET
Traditionally, foreign exchange settlement was carried out bilaterally between trade parties through the use of Chart 4 14 correspondent banking arrangements. Such arrangements Value Settled in CLS Bank Compared with Fedwire lead to exposures because there is no direct link between the and TARGET, 2003-06 payment of the two currency legs; thus, there is a risk (called Three-Month Moving Average Herstatt risk) of paying the currency sold but not receiving the Trillions of U.S. dollars currency bought. This risk, combined with the vast size of daily 80 FX trading and the global interdependence of FX markets and 70 payments systems, raised concerns among central banks. CLS Bank total In March 1996, the BIS issued a report titled “Settlement 60 TARGET Risk in Foreign Exchange Transactions.” The report analyzed 50 in particular the risks associated with FX settlement operations Fedwire and outlined a strategy for reducing them. Based on this report, 40 CLS Bank USD only the G-10 central banks endorsed a three-track strategy to 30 reduce these risks. First, individual banks were requested to 20 CLS Bank euro only 14 Correspondent banking is an arrangement under which one bank 10 (correspondent) holds deposits owned by other banks (respondents) and provides payment and other services to those respondent banks (Bank for 0 International Settlements 2003). Correspondent relationships are sometimes 2003 04 05 06 07 formalized; the benefit is that parties to a trade do not have to know each other because the system’s rules replace individual agreements between them. Sources: Bank for International Settlements; authors’ calculations.
FRBNY Economic Policy Review / September 2008 67 Table 2 Even though major progress has been made in reducing Total Foreign Exchange Obligations Settled, aggregate foreign exchange settlement exposures, the size by Method, 1997 and 2006 and duration of the exposures still settled by traditional correspondent banking are significant. In its assessment, the Percentage of Total Value CPSS recommends actions to address the remaining exposures Value (Trillions that may continue to present systemic risk. Settlement Method of U.S. Dollars) 1997 2006
CLS Bank (PvP) 2,091 0 55 Traditional correspondent banking 1,224 85 32 Bilateral netting 304 8 6.Trend 5: Increasing Settlement 15 Other 203 5 Values and Volumes Total 3,821 100 100 Available data suggest that the values transferred over large- Source: Bank for International Settlements (2007b). value payments systems in Committee on Payment and Note: PvP is payment-versus-payment. Settlement Systems countries grew substantially during the 1980s and 1990s. Table 3 summarizes the annual growth in nominal payment values in local currencies since 1985 for four systems for which such long-ranging data are available. In the United States, the number of transfers originated over both settled five times the euro value settled in CLS Bank, Fedwire Fedwire and CHIPS grew by 6 percent per year on average from settled about 1.7 times the USD value settled in CLS Bank. 1985 to 2000. In terms of value, turnover increased by an Ten years after the strategy was launched, the Committee on average of 9 percent per year. Since 2000, both systems have Payment and Settlement Systems conducted a survey to assess experienced a slight slowdown in growth, with volumes the extent to which systemic risk had been reduced (Bank for increasing by 4 percent and 5 percent and values by 7 percent International Settlements 2007b). The survey included 109 and 9 percent, respectively. institutions that represent 80 percent of the FX market in the fifteen currency areas.15 The institutions reported average daily gross values of FX settlement obligations totaling USD 3.8 trillion (Table 2). Of these obligations, 32 percent (USD 1.2 trillion) were settled by Table 3 Annual Turnover Growth for Selected Large-Value traditional correspondent banking arrangements and are still Payments Systems, Nominal Values subject to settlement risk. Nonetheless, this is a significant Percent improvement from the time of the 1997 survey, when an estimated 85 percent of the obligations were settled by this Average Annual Growth method. The major reason for the decline in FX obligations subject to settlement risk is the increasing use of CLS Bank. CHAPS Fedwire CHIPS SIC
In 2006, CLS Bank settled 55 percent (USD 2.1 trillion) of the Volume total FX settlement obligations of the surveyed institutions. Furthermore, 8 percent of FX obligations were settled by 1985-2006 14 6 6 11 bilateral netting.16 Other settlement methods, such as the PvP 1985-2000 16 6 6 9 2000-06 7 4 5 13 arrangement available in Hong Kong’s USD CHATS and EUR CHATS systems or on-us settlement,17 accounted for the Value remaining 5 percent of obligations. 1985-2006 17 8 8 2 15 The survey updated and extended previous CPSS surveys conducted in 1996 1985-2000 22 9 9 4 (Bank for International Settlements 1996) and 1997 (Bank for International 2000-06 3 7 5 0 Settlements 1998). 16 Provided it is conducted under legally robust arrangements, bilateral netting Sources: CHIPS; Federal Reserve Bank of New York; Bank of England; can also be a safe and efficient method for reducing settlement exposures (Bank Swiss National Bank; authors’ calculations. for International Settlements 2007b). 17 In on-us settlement, both legs of an FX trade are settled on the books of a Notes: Systems are described in Appendix A. For SIC, the growth rates single institution (Bank for International Settlements 2007b). are calculated using data from 1989 to 2006.
68 Global Trends in Large-Value Payments Chart 5 Chart 6 shows the relative importance of the CPSS Time Series of Value of Transfers in CHIPS countries’ interbank payments systems in terms of value and Fedwire since 1985 settled. Currently, the three largest systems represent more Trillions of U.S. dollars Trillions of U.S. dollars than 75 percent of value transferred. The six largest systems 600 90 Fedwire and CHIPS multiple of GDP represent almost 95 percent. The largest individual LVPS is Scale 500 75 TARGET, with an annual settlement value of EUR 489 trillion (USD 607 trillion). The two U.S. systems, Fedwire and CHIPS, 400 60 CHIPS respectively rank second and third. However, their combined Scale 300 45 share of turnover, 45 percent, exceeds the share of combined turnover of the euro payments systems TARGET and EURO1, 200 30 Fedwire 34 percent. Scale 100 15 The number of transfers settled increased in all of the CPSS countries’ payments systems (Chart 7). The Swiss 0 0 1985 90 95 00 05 SIC system stands out, with more than 250 million payments settled in 2005. SIC is notable because there is no separate Sources: CHIPS; Federal Reserve Bank of New York; authors’ system for settling retail payments in the Swiss payment calculations. infrastructure. Thus, retail as well as wholesale payments are settled in SIC. (The settlement of low-value payments in LVPSs
In the United Kingdom, CHAPS saw double-digit turnover growth in terms of volumes and values from 1985 to 2000. However, since the turn of the century, growth has decreased. In Switzerland, the picture is a bit different, as volumes have Chart 6 been growing faster since 2000. If growth is measured in Swiss Percentage of Transfer Value in CPSS Countries’ francs, though, turnover since 2000 has been almost flat, at Large-Value Payments Systems, 2005 around CHF 45 trillion.
Chart 5 shows the annual turnover for Fedwire and CHIPS EURO1 Others CHAPS $53 $121 since 1985. Fedwire turnover increased from around USD $94 (3 percent) (6 percent) 100 trillion in 1985 to more than USD 570 trillion in 2006. (5 percent) Values settled in CHIPS were smaller than those settled in Fedwire from 1985 to 1988. However, between 1988 and 1998, settlement values in CHIPS surpassed those in Fedwire. BoJ-NET After settlement values in CHIPS dipped in the late 1990s,18 $196 (10 percent) they have steadily risen again and amounted to USD 395 trillion in 2006. TARGET A major determinant of the value of interbank payments $607 is general economic activity. Since 1990, the combined CHIPS (31 percent) $350 settlement value on Fedwire and CHIPS has kept pace with (18 percent) economic activity, at around seventy times GDP (Chart 5, right axis). However, from 1985 to 1990, combined turnover on the two systems rose from forty-five times GDP to almost Fedwire $519 seventy-five times GDP. Likely explanations are technological (27 percent) development, deregulation of financial markets, and innovation in financial instruments.19
18 This point is addressed in more detail in our Trend 9 discussion. 19 6Similar developments also took place outside the United States. Annual Sources: Bank for International Settlements; authors’ calculations. turnover in the United Kingdom’s CHAPS system rose from seven times GDP in Notes: Dollar amounts are in trillions of U.S. dollars. “Others” includes SIC, its first full year of operation, 1985, to more than fifty times GDP at the turn of LVTS, PNS, HKD CHATS, K-RIX, MEPS, and USD CHATS (in size order). the century. In Switzerland, annual turnover reached 120 times GDP in 1997; Systems are described in Appendix A. CPSS is the Committee on Payment it has since fallen to around 90 times GDP in 2006, the same level as in 1989. and Settlement Systems.
FRBNY Economic Policy Review / September 2008 69 Chart 7 Chart 8 Transactions Settled by CPSS Countries’ Value of Average Payment in CPSS Countries’ Large-Value Payments Systems, Large-Value Payments Systems, 1999 and 2005 1999 and 2005
SIC BoJ-NET Fedwire K-RIX TARGET TARGET CHIPS LVTS EURO1 CHIPS CHAPS Sterling Fedwire PNS HKD CHATS BoJ-NET CHAPS Sterling LVTS MEPS HKD CHATS 2005 PNS 2005 MEPS 1999 EURO1 1999 K-RIX SIC 0 50 100 150 200 250 300 0 10203040506070 Millions of transactions per year Millions of U.S. dollar equivalent
Sources: Bank for International Settlements; authors’ Sources: Bank for International Settlements; authors’ calculations. calculations. Notes: Systems are described in Appendix A. CPSS is Notes: Systems are described in Appendix A. CPSS is the Committee on Payment and Settlement Systems. the Committee on Payment and Settlement Systems.
is discussed in more detail in the next section.) Fedwire settled right-hand tail. In other words, a small number of payments the second highest number of transactions, 132 million. In account for a large share of value. In Fedwire, 5 percent of the 2005, TARGET settled 72 million transactions and CHIPS largest payments account for 95 percent of the total value. settled 76 million. The considerable use of low-value wire transfers is not just a U.S. phenomenon. The distribution of payments handled in the Canadian LVTS is similar to that of the U.S system. The mean value is CDN 8 million (USD 6.6 million), while the 7.Trend 6: Shrinking Average median value is about CDN 50,000 (USD 41,300). In the Payment Sizes United Kingdom, the mean value of wire transfers processed by CHAPS Sterling is GBP 1.9 million (USD 3.45 million), and the Payments processed through large-value payments systems median value is estimated to be approximately GBP 25,000 come in many sizes. The maximum value of a payment allowed (USD 45,500). The bulk of payments in Switzerland’s SIC over Fedwire is one cent less than USD 10 billion; on occasion, system is less than CHF 5,000 (USD 4,000). payments of less than USD 1 are processed. In fact, most LVPSs Looking across large-value payments systems in CPSS process a significant amount of relatively low-value payments. countries, we observe a remarkable dispersion in average As a result, the dichotomy between small-value (retail) and payment value (Chart 8). In BoJ-NET, the average payment large-value (wholesale) payments systems is often blurred. In was in excess of 3.8 billion JPY (USD 35 million) in 2005—the Fedwire and CHIPS, the median payment size is less than USD highest among the LVPSs surveyed and three times larger than 35,000, and almost two-thirds of transfers are for amounts less the second-ranking K-RIX system of the Swedish Riksbank. than USD 100,000. Hence, both systems are important for Nonetheless, this is a significant drop from 1999, when the making low-value payments. The appeal of making low-value 20 transfers in Fedwire derives from the speed, certainty, and See Federal Reserve Board, “A Summary of the Roundtable Discussion on the Role of Wire Transfers in Making Low-Value Payments,” May 16, 2006 finality of settlement and, in some cases, from the ease of (available at
70 Global Trends in Large-Value Payments Chart 9 8.Trend 7: Falling Numbers Inflation-Adjusted Change in Average of System Participants Payment Size for CPSS Countries’ Large-Value Payments Systems, One of the most pervasive trends in international banking 1999-2005 is consolidation. All else equal, larger banks imply fewer participants in LVPSs. In addition, an increasing focus on costs K-RIX has made banks more selective in terms of the systems in which EURO1 they participate. It is no longer considered “a must” for many PNS foreign banks to clear their USD payments themselves as it was SIC in the 1980s and 1990s. However, technological advances have BoJ-NET CHAPS Sterling made it possible to participate in more systems in new ways, CHIPS as we describe in the emergence of cross-border and offshore LVTS systems (Trend 3). MEPS According to figures reported to the Bank for International Fedwire Settlements as part of its “Red Book” statistics, Fedwire had TARGET 10,000 participants in the late 1990s and 6,819 participants in HKD CHATS 2005. Despite this decrease, Fedwire is still by far the largest -100 -80 -60 -40 -20 0 20 40 LVPS in the CPSS group. The second largest system is Percent TARGET, with 2,628 participants. With only twelve direct Sources: Bank for International Settlements; authors’ participants, the Canadian LVTS is the smallest system in the calculations. CPSS group (Table 4). Notes: Systems are described in Appendix A. CPSS is Most LVPSs in the CPSS group saw participation decline the Committee on Payment and Settlement Systems. from 1999 to 2005 (Table 4). However, there was a substantial
average payment value was more than JPY 6,941 million (USD 63 million). The drop is attributable to the change from a deferred net settlement to a real-time gross settlement system Table 4 in 2001. An upper limit of JPY 5 billion per payment was set to Ranking of CPSS Countries’ Large-Value Payments Systems by Number of Participants, 1999 and 2005 allow for smoother settlement in the new RTGS system. TARGET is ranked third and followed by the North American systems. At the other end of the spectrum is the SIC system, Total Total Participants, Participants, Percentage with SFR 160,000 (USD 130,000), along with the euro systems System Ranking 1999 Ranking 2005 Change PNS (USD 2.9 million) and EURO1 (USD 1.2 million). Fedwire 1 9,994 1 6,819 -32 Adjusting for inflation, Chart 9 shows that the average TARGET 2 5,144 2 2,628 -49 payment size has fallen for most systems from 1999 to 2005. BoJ-NET 3 409 3 357 -13 For five of the twelve systems, it fell more than 40 percent. SIC 5 291 4 325 12 For Fedwire and TARGET, the average size has remained CHAPS 4 404 5 241 -40 unchanged. The only increase in payment size adjusted for HKD CHATS 6 151 6 129 -15 inflation occurred in the HKD CHATS system; it is to a large MEPS 7 136 7 111 -18 extent attributable to the period of deflation that Hong Kong EURO1 9 72 8 75 4 experienced between 1999 and 2004.21 CHIPS 8 77 9 48 -38 PNS 10 25 10 45 80 K-RIX 11 23 11 21 -9 LVTS 12 14 12 15 7
Sources: Bank for International Settlements; authors’ calculations.
21 The consumer price index in Hong Kong fell 12 percent between 1999 Notes: Systems are described in Appendix A. CPSS is the Committee and 2004. on Payment and Settlement Systems.
FRBNY Economic Policy Review / September 2008 71 increase in the number of participants in SIC and, even more synchronize with CLS Bank settlement hours during the so, in PNS.22 SIC’s growth in participation is attributable to its European morning. The CLS Bank settlement process takes opening to “remote membership” in 1998 (see our Trend 3 place during a five-hour window from 7:00 a.m. to noon discussion). The number of remote participants in SIC has since increased, whereas the number of Swiss participants has remained largely unchanged. A direct effect of the globalization of financial markets is the extension of operating hours in large-value 9.Trend 8: Extended Operating Hours payments systems.
A direct effect of the globalization of financial markets is the extension of operating hours in large-value payments systems. (CET). Its operating hours partially overlap the operating Between 1997 and 2005, there were two waves of operating hours of all the participating RTGS systems (Chart 10). hour extensions in CPSS countries’ large-value payments To meet industry requests to achieve greater overlap of U.S. systems. The first occurred in 1998 and 1999, when several wholesale payments system operating hours with those of the European LVPSs changed or extended their operating hours to Asia-Pacific markets, Fedwire and CHIPS expanded their coincide with TARGET’s business hours. The second wave operating hours in 1997 and 2004. In both instances, the coincided with the launch of CLS Bank in 2002. BoJ-NET, opening was moved to earlier while the closing remained MEPS, and LVTS extended their operating hours to unchanged. In 2007, SIC had the longest operating hours—23 hours and 22 For PNS, the increase in participants is attributable to an increase in the 15 minutes—approaching a twenty-four-hour settlement number of indirect participants; by and large, it can be explained by a reclassification of their status that took place over the period. cycle. The three North American systems—Fedwire, CHIPS,
Chart 10 Opening Hours of CLS Bank and Selected CPSS Countries’ Large-Value Payments Systems
BoJ-NET 2002
GMT + 9 MEPS 2001 2003 HKD CHATS EURO 1 TARGET
1997 K-RIX 1998 GMT + 1 TOP
(CET) PNS BI-REL 1999 ELLIPS 1999 SIC 2002 1999 GMT CHAPS 1998 2002 LVTS
2004 1997 CHIPS GMT - 5 (EST) 2004 1997 Fedwire CLS
0:00 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 Time
Sources: Bank for International Settlements; authors’ calculations. Notes: The blue and black shading represents the operating hours in 1996 or when the system was implemented. The full extent of all shaded areas (excluding the black areas, which represent opening hour shortenings in 1999 and before) is operating hours in 2005. Green shading indicates extensions in 1999 and before; grey shading shows extensions after 1999. Systems are described in Appendix A.
72 Global Trends in Large-Value Payments Chart 11 Most systems charge a fixed membership or admission fee Length of Operating Hours of Selected CPSS and the majority charge a per-transaction fee. However, Countries’ Large-Value Payments Systems, differences exist in terms of whether both the originator and 1997 and 2007 the receiver are charged for a transaction. TARGET only charges the originator whereas both Fedwire and SIC levy SIC fees on both the originator and the receiver. A simple, flat Fedwire transaction fee schedule is often used, but several systems CHIPS base the fee on a combination of the volume submitted by LVTS the participant, the value of the particular payment, the CHAPS Sterling submission time of the payment, and the mode of delivery, TARGET such as online and offline. In addition, participants may K-RIX BoJ-NET have to pay separate communications charges, for example, MEPS+ to SWIFT. CHATS 2007 In 1999, the Federal Reserve implemented a volume-based EURO1 1997 fee schedule to reflect more accurately the cost structure of PNS Fedwire services and its demand elasticity. This type of 0:00 3:00 6:00 9:00 12:00 15:00 18:00 21:00 0:00 structure remains in place today, in which offline participants Time are also assessed a surcharge to initiate or receive a funds Sources: Bank for International Settlements; authors’ calculations. Notes: Systems are described in Appendix A. CPSS is Large-value payments systems are the Committee on Payment and Settlement Systems. characterized by large economies of scale, as there are considerable fixed costs in terms of setting up and and LVTS—followed with 21.5, 20, and 17.5 hours, respec- tively. At the other end of the spectrum were the European maintaining the systems. In contrast, systems PNS and EURO1 and the Asian systems MEPS and individual payments generate diminutive HKD CHATS, with operating hours of between 8 and 9.5 hours. A reduction in operating hours occurred only in costs. This creates a potential problem the Singaporean system MEPS and, to a lesser extent, in the for efficient pricing. French PNS system (Chart 11).
transfer. In Switzerland, the originator of the payment is charged differently depending both on the time when the 10.Trend 9: Declining payment is submitted and its value. The SIC pricing schedule Transaction Fees is illustrated in Table 5. The receiver is charged a flat fee regardless of the settlement time and value. The fee structure Large-value payments systems are characterized by large provides an incentive for early input and settlement of economies of scale, as there are considerable fixed costs in payments, which in turn prevents the demand for settlement terms of setting up and maintaining the systems. In contrast, from peaking at the end of the day (see Heller, Nellen, and individual payments generate diminutive costs. This creates Sturm [2000]). For cross-border payments, TARGET has a potential problem for efficient pricing. Standard economic a transparent, volume-based pricing structure. Domestic theory suggests that each transaction should be priced at its payments are currently priced by each TARGET component marginal cost. However, marginal cost pricing implies that independently. the fixed costs are not recovered (see Holthausen and Rochet In the second half of the 1990s, the Federal Reserve [2006]). Central banks and system operators around the undertook a five-year project—akin to the current TARGET2 world have found different solutions to this challenge, initiative—to consolidate its processing facilities. The project depending on their mandate for the provision of payment resulted in significant savings that were passed on to users in services. the form of lower fees. The average transaction fee (nominal)
FRBNY Economic Policy Review / September 2008 73 Table 5 Pricing Principles in Fedwire, TARGET, and SIC Transaction Fee (U.S. Dollar Equivalent) Lowest Highest Mode of Volume-Based Value-Based Time of Day Delivery Sender Total Sender Total
Fedwire YNNY 0.1 0.2 0.3 0.6 TARGET YNNN 1.1 1.1 2.4 2.4
SIC NYYN 0.01 0.03 3.7 3.7
Sources: European Central Bank; Federal Reserve Bank of New York; Swiss National Bank.
Note: Total = sender + receiver, exchange rate of May 25, 2007.
in Fedwire dropped from USD 0.50 in 1996 to less than USD the trajectory of volumes processed over CHIPS declined 0.16 in 2006—a reduction of 68 percent (Chart 12). SIC significantly in the late 1990s and did not recover until 2001. At participants have seen a similar drop, as the average transaction the same time, Fedwire was experiencing steady growth. This fee fell from 17 centimes in 1999 to 7 centimes in 2005—a change coincides with the reduction in Fedwire fees and the reduction of 60 percent. TARGET transaction fees have been move to volume-based pricing. Conversely, in late 2005, fixed at the same nominal rates since 1999. CHIPS announced new incentive pricing for existing and new Pricing in LVPSs is important not only from a revenue and participants. Based on one year of data, incentive pricing cost perspective but also from a competitive perspective when appears to have had some effect, as Fedwire volume grew by private and public systems co-exist as they do in Europe and only 1 percent—the lowest rate in twenty years—while CHIPS the United States. As an example, the growth in the respective volume grew by 9 percent. volumes of payments submitted to Fedwire and CHIPS since 1990 is illustrated in Chart 12. Both systems have more than doubled the amount of payments they process. However, 11.Trend 10: Adoption of Common Standards for Large-Value Payments Systems Chart 12 The Effect of Fedwire Price Reductions In 1980, the governors of the central banks of the G-10 on CHIPS Volume countries established the Group of Experts on Payment Index: 1990 = 100 U.S. dollars per transaction Systems. One of the Group’s first projects was to conduct a 225 0.6 detailed review of payments system developments in the G-10 0.5 200 countries. It was published by the BIS in 1985 in the first of a series that has become known as the “Red Book.” The Group 0.4 175 also analyzed interbank netting schemes in the Angell Report 0.3 (“Report on Netting Schemes”) and in the Lamfalussy Report 150 CHIPS volume Average price (“Report of the Committee on Interbank Netting Schemes of Scale Scale 0.2 the Central Banks of the Group of Ten Countries”), published 125 0.1 by the BIS in 1989 and 1990, respectively. In 1990, the G-10 Fedwire volume a Scale Marginal price 100 Scale 0 governors established the Committee on Payment and 1990 92 94 96 98 00 02 04 06 Settlement Systems to assume and extend the activities of the Group of Experts on Payment Systems. The CPSS was set up as Sources: CHIPS; Federal Reserve Bank of New York; authors’ calculations. one of the BIS’ permanent central bank committees reporting a 80,000 or more transactions. to the G-10 governors.
74 Global Trends in Large-Value Payments The CPSS has focused on disseminating information on Estonia were assessed in 2000. In January 2001, the Governing payments system design and has been defining payment norms Council of the European Central Bank adopted the Core and best practices for the central bank community. In 1997, the Principles as the minimum standards for the Eurosystem’s CPSS published a report on real-time gross settlement systems common oversight policy on systemically important payments and in 2005 a report on new developments in large-value systems. As a consequence, the Governing Council decided that all nineteen SIPSs in the euro area would be assessed against the Core Principles in mid-2003. The standards published by the CPSS Exhibit 2 on the next page displays the countries that have [Committee on Payment and Settlement applied the Core Principles to assess their LVPSs. At the end of Systems] provide the main principles for 2006, fifty-nine countries had done so—up from twenty-two at the end of 2002. Of the fifty-nine countries, twelve had been the design and operation of payments both self-assessed and under the scrutiny of the IMF and World and settlement systems. Bank. In addition, payments systems in another forty-two countries had been assessed as part of an FSAP review, while five had conducted self-assessments. The United States belongs payments systems that focused on changes since the RTGS to the latter group. report. Furthermore, in January 2001, the CPSS published According to the assessments, practically all SIPSs in the “Core Principles for Systemically Important Payment developed countries meet the standards and codes in the Core Systems.” The document was developed to serve as guidelines Principles. This is also true for payments systems in many for promoting safety and efficiency in the design and operation emerging economies. Private sector payments systems in major of systemically important payments systems (SIPSs).23 currencies also comply with the Core Principles; notable The standards published by the CPSS provide the main examples are CHIPS and EURO1. principles for the design and operation of payments and settlement systems. They are currently being used as a reference by central banks and international organizations in their efforts to improve the safety and efficiency of payments systems 12.Future Developments worldwide. They are part of a set of key standards that the international community considers essential to strengthening This article describes ten long-range trends in the settlement of and preserving financial stability. As such, these standards are large-value payments. The questions worth considering are used by the joint IMF and World Bank Financial Sector how these trends will evolve and what new developments can Assessment Program (FSAP) and the Reports on the be foreseen. We offer some thoughts on these questions. Observance of Standards and Codes. Currently, the diffusion of RTGS is well under way. RTGS A growing number of LVPSs around the world have been and net settlement systems each have characteristics that make assessed according to the Core Principles. This has occurred them desirable, thus the hybridization of RTGS is likely to either as a self-assessment by the system operator, the central continue as long as liquidity is costly. Many central banks bank, or as part of an FSAP review. In some cases, systems have require collateral for intraday credit. With the ongoing even been self-assessed and assessed by the IMF and the World development of financial markets, collateral is likely to find Bank. Some central banks, such as the Bank of England and the new, more profitable uses than payment settlement. This will Swedish Riksbank, use the Core Principles for annual likely drive the cost of liquidity up and, as a consequence, assessments of their SIPSs as part of their payments system increase the demand for liquidity saving that netting and oversight. offsetting in conjunction with RTGS can offer. The trend The first country to assess its payments system according toward greater hybridization of systems is therefore likely to the Core Principles was Canada. In late 1999, Canada to continue. participated in an FSAP pilot that included an assessment of the The introduction of cross-border systems has been Canadian payments system, LVTS, using a draft version of the associated with unique events linked to the introduction of the Core Principles. The payments systems in Cameroon and euro and the establishment of CLS Bank. Cross-border systems are likely to remain rare in the future. However, remote 23 A payments system is systemically important when, if the system were participation may become more prevalent. insufficiently protected against risk, disruption within it could trigger or transmit further disruptions among participants or systemic disruptions in Offshore systems that settle a foreign currency are presently the financial area more widely (Bank for International Settlements 2003). small and serve niche markets—mainly a local FX market or
FRBNY Economic Policy Review / September 2008 75 Exhibit 2 Assessment of Compliance with the “Core Principles for Systemically Important Payments Systems,” 2006
No assessment Self-assesment FSAP Self and FSAP
Sources: International Monetary Fund (IMF); websites of countries’ central banks. Note: FSAP is the joint IMF-World Bank Financial Sector Assessment Program.
the needs of banks in the area and time zone to settle payments deregulation and innovation in the 1980s and early 1990s has in a foreign currency among each other. Such demands may largely been absorbed. arise in the context of the establishment of new financial The average real value of payments processed in LVPSs has centers, for instance, in the Middle East or China, where the declined. As transaction prices seem to be declining too, it can People’s Bank of China is developing a USD clearing system. be expected that the benefits of real-time settlement will Most existing or planned offshore systems are limited to a outweigh the costs for a wider variety of smaller financial single country. With improvements in information and transactions. Thus, we expect that the average value of large- communications technologies, the fixed cost of setting up such value payments will continue to fall. systems is being reduced. As a consequence, we may see more Consolidation in financial services is continuing. Especially offshore systems emerge, but they are likely to remain niche in Europe, the process of cross-border mergers has not yet players, much like the existing ones are. taken off. In addition, the introduction of TARGET2 and the CLS Bank settlements have grown rapidly and are likely consolidation of all the EU RTGS systems into a single entity to continue to do so. Currently, CLS Bank captures around will substantially reduce the number of LVPS participants, as 50 percent of all FX trades and is pursuing the settlement of banks operating in several EU countries will be better other types of transactions. As most of its costs stem from fixed positioned to manage their payments centrally. investments, CLS Bank has incentives to continue fostering One reason for the emergence of offshore systems is that the growth in settlement volumes. operating hours of domestic LVPSs do not coincide with the Settlement values are likely to continue growing at the pace business hours of LVPSs in other countries. Often, the of GDP in the long run, and be cyclical to financial market operating hours of euro or U.S. dollar LVPSs do not sufficiently activity in the short run—as they have done over the past ten overlap those of Asian financial centers. Fedwire and CHIPS years. The rapid growth in values attributable to financial have reacted to this disparity by extending their operating
76 Global Trends in Large-Value Payments hours. The European systems have had less of a need to do so technology, and perhaps competition between the public and because their operating hours already overlap those of Asian private systems that operate side by side in some countries. financial centers by a few hours. The overlap with current These reasons are not likely to change, and the cost of making operating hours, however, is now wider for the U.S. dollar payments is likely to continue to fall. than for the euro—in spite of the more advantageous time The final trend we discussed was the standardization of differences in Europe. large-value payments systems through the use of common Evidence from systems for which price data are available standards. The “Core Principles for Systemically Important suggests that the cost of payments in LVPSs has declined Payment Systems” is already widely accepted and will continue rapidly. The underlying reasons are associated with regulatory to be applied around the world. changes, lower costs of information and communications
FRBNY Economic Policy Review / September 2008 77 Appendix A: Large-Value Payments Systems in CPSS Member Countries
Annual Value Annual Number of Transactions Year of of Transactions (2005, in Billions Country System Name Abbreviation Implementation (2005, in Thousands) of U.S. Dollars)
Belgium Electronic Large-Value Interbank Payments System ELLIPSa 1996 1,800 21,448 Canada Large-Value Transfer System LVTS 1999 4,600 30,321 France Transferts Banque de France TBFa 1997 4,300 151,425 France Paris Net Settlement PNS 1999 6,800 19,432 Germany RTGSplus RTGSplusa 2001 35,800 172,023 Hong Kong HK Dollar Clearing House Automated Transfer System HKD CHATS 1996 4,100 14,936 Hong Kong U.S. Dollar Clearing House Automated Transfer System USD CHATS 2000 1,500 1,588 Hong Kong Euro Clearing House Automated Transfer System Euro CHATS 2003 1,000 422 Italy BI-REL BI-REL1 1997 10,400 40,840 Japan BoJ-NET Funds Transfer System BoJ-NET 1988 5,300 196,452 Netherlands TOP TOPa 1997 4,700 38,126 Singapore Monetary Authority of Singapore (MAS) Electronic Payments System MEPS 1998 2,500 7,564 Sweden K-RIX K-RIX 1990 1,500 14,867 Sweden E-RIX E-RIXa 1999 1,000 279 Switzerland Swiss Interbank Clearing SIC 1987 256,400 32,956 Switzerland EuroSIC EuroSIC 1999 9,130 630 United Kingdom CHAPS Sterling CHAPS Sterling 1984 29,600 94,299 United Kingdom CHAPS Euro CHAPS EURO1 1999 5,100 66,859 United States Fedwire Funds Service Fedwire 1918 123,400 518,547 United States Clearing House Interbank Payments System CHIPS 1970 71,500 349,871 European Union Trans-European Automated Real-time Gross settlement Express Transfer system TARGET 1999 76,200 607,254 European Union EURO1 EURO1 1999 46,400 53,334 European Union ECB Payment Mechanism EPMa 1999 41 5 International Continuous Linked Settlement Bank CLS Bankb 2002 47,900 785,300
Notes: In this article, we analyze the European TARGET system but do not consider its components (ELLIPS, TBF, RTGSplus, BI-REL, and TOP). CPSS is Committee on Payment and Settlement Systems.
a The system is a component of TARGET, which consists of fifteen national real-time gross settlement systems and the EPM system of the European Central Bank. b CLS Bank data are based on the aggregation of both sides of a foreign exchange transaction.
78 Global Trends in Large-Value Payments Appendix B: TARGET and TARGET2
TARGET (Trans-European Automated Real-time Gross Share of Values Transferred by TARGET Components settlement Express Transfer system) is the RTGS system for the euro, owned and operated by the Eurosystem. It went live in January 1999. The system is used for the settlement of central bank operations, large-value euro interbank transfers, and Others other euro payments. TARGET was created by interconnecting (19 percent) national euro RTGS systems and the European Central Bank ELLIPS RTGSplus (4 percent) payment mechanism. (28 percent)
Within TARGET, the two largest components—the TOP German RTGSplus and the French TBF—process more than (6 percent) half of the values transacted. The next three largest systems— the United Kingdom’s CHAPS Euro, the Italian BI-REL, and TBF the Dutch TOP—process another quarter of transfer values, BI-REL (25 percent) (7 percent) while the remaining ten smaller components are responsible for the remaining quarter or less. The decentralized infrastructure of TARGET is being CHAPS Euro replaced by TARGET2, which is based on a single technical (11 percent) platform. TARGET2 went live in November 2007. All Eurosystem central banks and Kronos Euro will participate in Source: European Central Bank, TARGET 2006 Annual Report. the new system. E-RIX discontinued operations on January 1, 2007, and CHAPS Euro will not connect to TARGET2.
FRBNY Economic Policy Review / September 2008 79 Appendix C: Cross-Border Payments in TARGET
In 2006, TARGET processed on average 326,000 payments a This simply reflects the importance of London as a financial day worth about ¤ 2.1 trillion. Inter-member state or cross- center. Interestingly, euro payment flows within the United border payments accounted for 23 percent of the volume and Kingdom are minuscule in comparison. In contrast, the value 35 percent of the value of payments transferred (see table). of cross-border payments to and from Italy, Spain, and Not surprisingly, the cross-border flow of payments within Portugal is less than the countries’ respective shares of EU gross TARGET correlates with economic output and the size of the domestic product. The opposite is true for the Benelux financial sectors across countries. The geographical network of countries, all of which host important financial centers payment flows in Europe is shown in the exhibit. Data are from (light blue lines). 1999-2002 and hence exclude some of the newer accession countries. The size of each node is proportional to the value of intra-border or domestic payment flows, and the width of Geography of TARGET Cross-Border Payments links between countries is proportional to the value of cross- Finland border flows. The largest flows are between the three largest Sweden economies: Germany, the United Kingdom, and France. Although the United Kingdom has not yet adopted the euro, it has a prominent role in cross-border euro payment flows. Belgium Ireland Netherlands Great Britain Germany
lnter- and Intra-Member State Payments Luxembourg in TARGET, 2006 France
Value Volume
Daily Daily Average Average
(Billions Share (Thousands Share Italy
of Euros) (Percent) of Payments) (Percent) Spain
Portugal Total 2,092 326 Greece Inter-member state 725 35 75 23 Intra-member state 1,368 65 252 77
Source:
80 Global Trends in Large-Value Payments References
Arjani, N., and D. McVanel. 2006. “A Primer on Canada’s Large-Value Bech, M., and K. Soramäki. 2001. “Liquidity, Gridlocks, and Bank Transfer System.” Bank of Canada. Failures in Large-Value Payments Systems.” Central Banking Publications, E-Money and Payments System Review: 111-26. Bank for International Settlements. 1996. “Settlement Risk in Foreign Exchange Transactions.” Committee on Payment and Settlement ———. 2005. “Systemic Risk in a Netting System Revisited.” Systems, Publication no 17. Basel, Switzerland. In H. Leinonen, ed., Liquidity, Risks, and Speed in Payment and Settlement Systems—A Simulation Approach, 275-96. ———. 1998. “Reducing Foreign Exchange Settlement Risk: Bank of Finland Studies in Economics and Finance. A Progress Report.” Committee on Payment and Settlement Systems, Publication no 26. Basel, Switzerland. Heller, D., T. Nellen, and A. Sturm. 2000. “The Swiss Interbank Clearing System.” Swiss National Bank, Payment System ———. 2001. “Core Principles for Systemically Important Payment Subsection, June. Available at
———. 2007a. “Statistics on Payment and Settlement Systems in Leinonen, H., and K. Soramäki. 1999. “Optimizing Liquidity Usage Selected Countries—Figures for 2005.” Committee on Payment and Settlement Speed in Payment Systems.” Bank of Finland and Settlement Systems, Publication no 78. Basel, Switzerland. Discussion Paper no. 16, December.
———. 2007b. “Progress in Reducing Foreign Exchange Settlement McAndrews, J., and J. Trundle. 2001. “New Payment System Designs: Risk.” Committee on Payment and Settlement Systems, Causes and Consequences.” Bank of England Financial Publication no 81. Basel, Switzerland. Stability Review 11, December: 127-36.
Bech, M. 2007. “The Diffusion of Real-Time Gross Settlement.” Rosati, S., and S. Secola. 2006. “Explaining Cross-Border Large-Value In S. Millard, A. Haldane, and V. Saporta, eds., The Future Payment Flows— Evidence from TARGET and EURO 1 Data.” of Payments. Routledge International Studies in Money and Journal of Banking and Finance 30, no. 6 (June): 1753-82. Banking. London: Routledge. Selgin, G. 2004. “Wholesale Payments: Questioning the Market- Bech M., and B. Hobijn. 2007. “Technology Diffusion within Central Failure Hypothesis.” International Review of Law and Banking: The Case of Real-Time Gross Settlement.” Economics 24, no. 3 (September): 333-50. International Journal of Central Banking 3, no. 3 (September): 147-81.
FRBNY Economic Policy Review / September 2008 81
Olivier Armantier, Jeffrey Arnold, and James McAndrews
Changes in the Timing Distribution of Fedwire Funds Transfers
• The Federal Reserve’s Fedwire funds transfer 1.Introduction service has long displayed a concentrated peak of activity in the late afternoon. he Federal Reserve’s Fedwire funds transfer service is the Tbiggest large-value payments system in the United States • Sending large payments late in the day can in terms of participants, value, volume, and use by other heighten operational risk by increasing the settlement systems. Although Fedwire funds activity has long potential magnitude of liquidity dislocation been concentrated in the late afternoon, recently there has been and risk if operational disruptions occur. a noticeable shift to later in the day. The value of funds activity after 17:00 has increased from 20 percent in 1998 to more than • A study of the distribution of Fedwire 30 percent in 2005 (Board of Governors of the Federal Reserve payments finds that the peak of the timing System 2006). In 2006, the Federal Reserve commented on the distribution has become more concentrated, risk posed by this change: has shifted to later in the day, and has been “From an operational risk perspective, delaying the divided into two peaks. sending of large payments until late in the day increases the potential magnitude of liquidity dislocation and risk • These trends are likely explained by a higher in the financial industry if late-in-the-day operational value of payments transferred, the settlement disruptions should occur. An increase in such risk is patterns of private settlement institutions, particularly troublesome in an era of heightened concern and increased industry concentration. about operational disruptions from a range of sources” (Board of Governors of the Federal Reserve System 2006). The study uncovers no specific evidence • There is a complex set of trade-offs between risks and costs of heightened operational risk associated in large-value payments systems (Bank for International with late activity, but it points to a high level Settlements 2005). Theory suggests that the concentration of of interaction between Fedwire and private late-afternoon Fedwire activity is the result of coordination settlement institutions. among banks to reduce liquidity costs, delay costs, and credit risks. As these costs and risks change over time, we would
Olivier Armantier is a senior economist at the Federal Reserve Bank of The authors thank the managements of CHIPS and DTC for permission to New York; Jeffrey Arnold was a research associate at the Bank at the time this report data associated with their operations. They also thank Craig Furfine article was written; James McAndrews is a senior vice president at the Bank. and other seminar participants at the Federal Reserve Bank of Chicago and the
FRBNY Economic Policy Review / September 2008 83 expect the timing of payment activity to be affected. In this of the amount of Fedwire value submitted by different system article, we seek to quantify how the changing environment in participants). which Fedwire operates has affected the timing of payment Our study proceeds as follows. The next section reviews the value transferred within the system. literature on models of banks’ payment timing decisions.2 We observe several trends in payment timing from 1998 to Section 3 considers how Fedwire payments are currently 2006. After 2000, the peak in payment activity shifts to later distributed across the time of day and how the distribution has in the day. Indeed, post-2000, a greater concentration of changed over time. In Section 4, we use regression analysis to payments occurs after 17:00. At the same time, however, several examine which factors have been most relevant in explaining factors have been associated with increased payment activity these trends. Section 5 focuses on the influence that settlement early in the day, such as the creation of the Continuous Linked institutions have on the timing of Fedwire payments. We Settlement (CLS) Bank, an institution that settles U.S. dollar conclude with a brief summary of our observations and offer payments early in the morning; changes to the Clearing House suggestions for further research. Interbank Payments System’s (CHIPS) settlement practices; and expanded Fedwire operating hours. Despite these developments, we find that the distribution of payment activity across the day still peaks more in the late afternoon. 2. Literature Review Payments made through Fedwire are distributed through- out the operating day, so no single statistic fully captures Banks seek to minimize the costs associated with sending the changes in timing that we observe. To analyze the timing payments. The existing models of banks’ payment-sending behavior generally focus on four factors: the cost of liquidity, We seek to quantify how the changing the cost of delay, uncertainty (both strategic and structural) and settlement risk, and the instruction arrival process. environment in which Fedwire operates The cost to banks of settling a payment is a function of has affected the timing of payment value the cost of liquidity used to send the payment and the cost transferred within the system. associated with delaying it. The cost of liquidity is the cost a bank faces when using account balances (usually modeled as being held at the central bank) in a payments system. This cost of payments on Fedwire, we measure the times at which is often zero as long as the bank has a positive balance in its each percentile of value transferred on a particular day was account and nonzero if the bank needs to borrow money to completed. In addition, we use regression analysis to examine settle the payment. In Fedwire, the Federal Reserve provides factors associated with the intraday timing of each percentile of intraday (daylight) credit in the form of overdrafts for which value transferred by Fedwire over time. Explanatory variables banks are charged a marginal fee, if the bank is eligible for here include changes in the Federal Reserve’s Payments System daylight credit and has exceeded the amount of credit that falls Risk Policy and the activity of settlement systems. We measure within a deductible amount (Board of Governors of the Federal the effects of multiple explanatory variables on the whole time Reserve System 2004). Many other payments systems provide distribution of payments to understand more fully why parts collateralized intraday liquidity at zero marginal cost after the of the Fedwire timing distribution have changed. bank posts collateral. An additional source of liquidity in Our study focuses on two notable changes that have affected payments systems is the overnight money market for federal the higher percentiles of value transferred on Fedwire: the funds and repos. The cost of delay is the compensation a bank 60th-90th percentiles. First, the most concentrated period of must pay a customer for a delayed payment or the reputational value transfer on Fedwire has moved to later in the day, from damage and loss of future business a bank suffers from delaying 16:48 to 17:11; second, the concentration of Fedwire value customer payment requests (Angelini 1998). transferred has increased. Together, these changes have Angelini (1998, 2000) considers the behavior of banks vis-à- 1 resulted in greater percentages of value transferred after 17:00. vis liquidity and delay costs in a real-time gross settlement Our analysis suggests that these changes are explained largely (RTGS) payments system. He shows that the equilibrium in an by the change in the timing of CHIPS end-of-day settlement RTGS system involves excessive delay of payments, as banks do activity, growth in the volume and value of Fedwire payments, not fully internalize the benefits to other banks from the receipt and growth in the pattern of industry concentration (a measure 2 For simplicity of exposition, we use the term banks to refer to direct 1 We measure this peak as the median daily time of the top ten minutes of participants in the payments system, although these participants may not Fedwire value. necessarily be banks.
84 Changes in the Timing Distribution of funds. Bech and Garratt (2003) find that RTGS systems can addition, the number of banks in a payments system—whether be characterized by multiple equilibria, with the relative costs 10 or 1,000—can influence the likelihood of successful of liquidity and delay determining whether the system has an coordination of payments. early or late equilibrium. Additionally, whether intraday A number of studies provide empirical evidence on liquidity liquidity is provided as priced or collateralized credit influences costs. McAndrews and Rajan (2000) document the payment the type of “game” and the associated equilibria that would and value timing distributions on the Fedwire funds service. result. Other papers that consider how changes in liquidity McAndrews and Potter (2002) show the effects of the terrorist attacks of September 11, 2001, on the timing of payments over Fedwire. Mills and Nesmith (2008) find that the charging of An observed late-day distribution of overdraft fees on Fedwire in 1994 sped up the settlement of payments could occur either because payments on the securities service but did not change the timing of payments on the funds service. Heller, Nellen, and Strum banks delay payments or because banks (2000) show that the introduction of intraday credit to banks receive payment instructions late in in the Swiss Interbank Clearing system dramatically shifted the value time distribution of payments to earlier in the day. the day. McAndrews (2006) finds that high-payment-value days lead to later value-weighted average payment value settlement times on Fedwire, consistent with a model of higher shadow prices of prices affect payment timing are Bech (2008) and Mills and liquidity on high-payment-value days. Becher, Galbiati, and Nesmith (2008). Green (2005) discusses the welfare Tudela (2008), examining the timing of sterling payments on implications of these models and questions whether there are the Clearing House Automated Payment System (CHAPS), find social costs to delay. that CHAPS has a less pronounced late-in-the-day peak than Kahn, McAndrews, and Roberds (2003), Mills and Nesmith the Fedwire funds service does. The divergent patterns in (2008), and Bech (2008) focus on settlement risk in payments Fedwire and CHAPS are broadly consistent with the models systems. Settlement risk is the uncertainty to which banks are of Bech and Garratt (2003) and Bech (2008), but they also subject when they face the choice of submitting or delaying likely reflect the imposition and maintenance of throughput a payment. Participant A may expect an offsetting payment guidelines by CHAPS participants; the guidelines govern the from Participant B to occur later in the day. However, percentage of value to be submitted at different times during the uncertainty about whether Participant B might either default day. Finally, several papers use simulations to examine the or delay sending payment until the next day can result in trade-off between liquidity costs and delays in theoretical Participant A’s decision to delay delivery of its payment to payments systems, including Koponen and Soramäki (2005) the RTGS system. and Leinonen and Soramäki (2005). An important but often overlooked assumption of these models is the time at which customers submit payment instructions to their bank. It is only after a bank receives an instruction from its customer (including the bank itself) that 3. Descriptive Analysis the bank decides whether to send settlement instructions to the payments system or delay settlement. An observed late-day Here we analyze payment and value time distributions on the distribution of payments could occur either because banks Fedwire funds service. We focus on the number and value of delay payments or because banks receive payment instructions payments transferred during a minute and contrast the time late in the day. For example, after CLS Bank began operations, series of these variables in 1998 and 2006. Our work is purely banks had a new stream of payments to submit at a particular descriptive; in the next section, we conduct regression analysis time of day, which can cause significant changes in the overall exploring the reasons behind the changes observed. value time distribution of payments. For our examination, we remove all payments to or from the A factor that has not been discussed in the literature but may settlement institutions: CHIPS, CLS Bank, and the Depository influence payment timing decisions is the industrial structure Trust Company (DTC). This allows us to focus on the non- of banks that participate in the payments system. Industrial settlement institutions’ funds transfers on Fedwire, as these are structure can differ for a number of reasons, many of which subject to the strategic decisions of the sending party. Notably, may be exogenous to banks’ activities in the payments system. this approach excludes the early-morning activity of CHIPS Such differences can result in varying costs of liquidity. In and CLS Bank.
FRBNY Economic Policy Review / September 2008 85 3.1 General Pattern in the Timing of Payments service. After 08:30, payment volume initially declines, then steadily rises to a plateau between 15:00 and 16:00, when it Charts 1 and 2, respectively, present the probability distri- reaches a level of activity similar to that of the 08:30 period. bution functions for the percentage of the daily number of After 16:00, payment volume drops sharply until the close transfers and the percentage of daily value settled. Each point of Fedwire. on Chart 1 (Chart 2) represents the average number (value) The value of payments settled also peaks in the late of payments transferred during that minute expressed as a afternoon, but the peak is both sharper and later than that of percentage of the total number (value) of payments transferred the volume of payments (Chart 2). Like volume, value settled that day. The charts show that the timing of Fedwire payments exhibits a general pattern that remained essentially stable between 1998 and 2006, whereby both the number and value of The timing of Fedwire payments exhibits payments peaked in the late afternoon. We start by describing a general pattern that remained essentially this general pattern. stable between 1998 and 2006, whereby Chart 1 shows that relatively few payments are sent before 08:00, with the notable exception of the period following the both the number and value of payments opening of Fedwire (00:30 in 1998 and 21:00 in 2006), when peaked in the late afternoon. many banks submit their CHIPS prefunding payments (recall that funds transfers to or from CHIPS and to or from CLS Bank are excluded from our measure of funds transfers in these is low during the early-morning hours and it rises to its peak charts). In the ten-minute period beginning at 08:30, there is a around the time DTC and CHIPS close. Note that while large spike in payment activity in 1998 and in 2006. The spike payment volume is falling after 16:00, payment value is rising. partially results from increases in bank balances attributable to This is the result of the concentration of large-value payments the Federal Reserve Bank of New York’s making principal and late in the day. Finally, although discernible, the peaks around interest payments on behalf of the U.S. Treasury and the opening of Fedwire and at 08:30 are proportionally much government-sponsored enterprises (GSEs) as well as customer lower for the value of payments than for the volume of activity associated with the opening of the Fedwire securities payments.
Chart 1 Fedwire Funds Payment Time Distribution, 1998 and 2006
Percentage of daily number of payments 0.30
1998 0.25
0.20 2006 0.15
0.10
0.05
0 21 22 23 0 1 234 5 678 9 10 11 12 13 14 15 16 17 18 Hour
Sources: Federal Reserve Bank of New York; authors’ calculations. Notes: The chart shows the mean daily percentage of total payments settled in each minute. Values exclude payments associated with CHIPS, CLS Bank, DTC, and principal and interest payment funding.
86 Changes in the Timing Distribution Chart 2 Fedwire Funds Value Time Distribution, 1998 and 2006
Percentage of daily value of payments 1.00 2006
0.75
0.50
1998
0.25
0 21 22 23 0 1 234 5 678 9 10 11 12 13 14 15 16 17 18 Hour
Sources: Federal Reserve Bank of New York; authors’ calculations. Notes: The chart shows the mean daily percentage of total payments settled in each minute. Values exclude payments associated with CHIPS, CLS Bank, DTC, and principal and interest payment funding.
3.2 Changes in the Distribution of Value in 2006 than in 1998. The remaining deciles—30-90—moved to later-day settlement. Additionally, there is decreased A comparison of Charts 1 and 2 indicates that the value time variability in the settlement time of deciles 40-100, implying distribution has changed much more than the volume time much more regularity in the later part of the value time distribution. Therefore, we now focus on the evolution of the distribution. value time distribution. Table 1 presents the times at which the To understand the significance of these value shifts, we deciles of payment value settled on average in 1998 and 2006. calculate the time at which each percentile of daily value settles The earliest and latest deciles—10, 20, and 100—settled earlier for each day in 1998 and 2006. Then, for each percentile, we compare the distributions of the samples collected in 1998 and 2006 nonparametrically using the Mann-Whitney two-sample Table 1 statistic. Our results appear in Chart 3. As the x-axis indicates, Percentiles of Value Time Distribution each of the 100 points corresponds to a percentile, moving left to right from percentile 1 to percentile 100. The y-axis 1998 2006 corresponds to the number of minutes that must be added to Standard Standard Difference or subtracted from the 2006 sample until the Mann-Whitney Percentile Mean Deviation Mean Deviation (Minutes) test is insignificant at the 5 percent level. For instance, the first 10 11:15:21 00:09:34 10:37:37 00:15:22 -37.7 point at the bottom of Chart 3 indicates that the first percentile 20 12:58:33 00:14:09 12:44:10 00:18:02 -14.4 in 2006 settled significantly earlier than the first percentile in 30 14:24:21 00:15:41 14:33:51 00:17:60 9.5 1998 by at least 7 minutes and 30 seconds. However, a point on 40 15:19:59 00:14:58 15:38:09 00:13:25 18.2 the origin line indicates that the corresponding percentile 50 15:59:29 00:14:33 16:15:04 00:10:44 15.6 60 16:28:18 00:12:09 16:37:29 00:08:22 9.2 cannot be distinguished statistically between the two samples. 70 16:46:05 00:10:46 16:58:58 00:07:09 12.9 Percentiles 1-24 are located below the origin axis, indicating 80 17:02:47 00:10:43 17:14:06 00:04:20 11.3 that the lower percentiles of value were transferred earlier in the 90 17:26:25 00:11:27 17:30:12 00:03:53 3.8 day in 2006 relative to 1998. Percentiles 27-95 lay above the 100 18:37:28 00:23:05 18:31:14 00:07:15 -6.2 origin line, indicating that most of the value distribution shifted Sources: Federal Reserve Bank of New York; authors’ calculations. to later in the day between 1998 and 2006. However, the points
FRBNY Economic Policy Review / September 2008 87 Chart 3 Mann-Whitney U Test on Percentiles of Value Time, 1998 to 2006
Minutes difference between 2006 and 1998 8
6
4
2
25 96 0
-2
-4
-6
-8 0 20 40 60 80 100 Percentile of time of value
Sources: Federal Reserve Bank of New York; authors’ calculations. Note: Minutes are subtracted from/added to each percentile until a Mann-Whitney rank-sum test is insignificant at the 5 percent level.
corresponding to the last percentiles, 96-99, are located below chart shows a clear positive time trend between 1998 the origin line. It appears that the last percentiles of value were and 2006. If we interpret the kurtosis as a measure of transferred later in the day in 1998. This result may be explained peakedness, Chart 4 confirms that the distribution of by a decline in the number and length of extensions since 1998. value settled has become more concentrated over time. The probability density function of the value time This result is important from a policy perspective, as it reflects a greater coordination in the timing of payments distributions in 1998 and 2006 is presented in Chart 2. Observe among Fedwire participants. A by-product of the greater first that the supports of the 1998 and 2006 distributions in the coordination of payment activity, the amount of value transferred on Fedwire that is offsetting within a ten- minute period rose significantly, from 56 percent in 1998 The highest peak of the distribution shifts to 58 percent in 2006. As the amount of offsetting to a later time, from around 16:48 in 1998 payments rises, banks enjoy greater economy in the use of liquidity. to around 17:11 in 2006. 2. The highest peak of the distribution shifts to a later time, from around 16:48 in 1998 to around 17:11 in 2006. In other words, the minute during which most of the daily chart have different lower bounds. Indeed, Fedwire operating value is transferred is now twenty-three minutes closer to hours were expanded when the opening hour was moved from closing time. Chart 5 shows the distribution of value 00:30 to 21:00 on May 17, 2004. Chart 2 shows that the change settled by time for each year between 1998 and 2006. It is in the opening hour did not dramatically affect the distribution clear from the chart that the highest peak moved to a later of value settled prior to 07:00. After 07:00, the two distributions time between 1999 and 2000, and it was not a gradual intersect several times, suggesting no clear pattern to how move. The mean daily time of the top ten contiguous the timing of payments changed between 1998 and 2006. We minutes of Fedwire funds value moved twenty minutes identify five distinctive features from our analysis of Chart 2: later, from 16:48 to 17:08, and the median daily time moved twenty-three minutes later, from 16:48 to 17:11, 1. The distribution of value settled becomes more supporting the presence of a significant shift toward a concentrated. In particular, observe that the magnitude later time in the peak of the distribution.3 of the highest peak is greater in 2006 than in 1998. To confirm this observation, we plot in Chart 4 the evolution 3 This difference is statistically significant at the 1 percent level for the Mann- of the kurtosis of the payment value distribution. The Whitney two-sample statistic.
88 Changes in the Timing Distribution Chart 4 Kurtosis of Fedwire Funds Value Time Distribution
January 18, 2000: January 22, 2001: December 8, 1997: CHIPS CHIPS September 10, 2002: May 17, 2004: July 1, 2006: Kurtosis Opening at 00:30 Closing at 17:00 Intraday Finality CLS Bank Opens Opening at 21:00 GSE Policy 4
3
2
1
0 1997 98 99 00 01 02 03 04 05 06
Sources: Federal Reserve Bank of New York; authors’ calculations. Notes: Kurtosis is the excess kurtosis. A twenty-one-day centered moving average is used. Values exclude payments associated with CHIPS, CLS Bank, DTC, and principal and interest payment funding. GSE is government-sponsored enterprise.
Chart 5 Fedwire Funds Value Time Distribution by Year
Percentage of total daily value 1.00 1998 1999 2000 0.75 0.50 0.25 0 21 23 1 357 9 11 13 15 17 21 23 1 357 9 11 13 15 17 21 23 1 357 9 11 13 15 17
1.00 2001 2002 2003 0.75 0.50 0.25 0 21 23 1 357 9 11 13 15 17 21 23 1 357 9 11 13 15 17 21 23 1 357 9 11 13 15 17
1.00 2004 2005 2006 0.75 0.50 0.25 0 21 23 1 357 9 11 13 15 17 21 23 1 357 9 11 13 15 17 21 23 1 357 9 11 13 15 17 Hour Hour Hour
Sources: Federal Reserve Bank of New York; authors’ calculations. Notes: The panels show the mean daily percentage of total payment value settled in each minute. Values exclude payments associated with CHIPS, CLS Bank, DTC, and principal and interest payment funding.
FRBNY Economic Policy Review / September 2008 89 3. At the time of highest activity (between 16:00 and 17:30), distribution of value settled for each year between 1998 the 2006 distribution exhibits three distinct peaks: two of and 2006 indicates that the 08:30 peak increased gradually comparable magnitude at precisely 16:30 and 17:00 and over time, while the 08:00 peak is present only in 2006. the third (the sharpest) at around 17:11. By comparison, The 08:00 peak is likely associated with the Federal the 1998 distribution possesses only two peaks: the Reserve’s July 2006 change to its Payments System Risk sharpest at around 16:45 and a slightly smaller peak at Policy regarding GSEs, while the 08:30 peak is likely 16:30. An analysis of the distribution of value settled for associated with the increased importance of the securities each year between 1998 and 2006 indicates that: 1) the markets and the opening of the Fedwire securities 16:30 peak is present every year throughout the sample service at 08:30. period; 2) as documented above, the highest peak moved from 16:48 to 17:11 between 1999 and 2000; and 3) the We conclude this section with an analysis of Chart 6, in emergence of the 17:00 peak can be traced to 2004. which the deciles of daily value settled are presented as a time series spanning the period from September 1997 to February 4. The 1998 distribution of value settled exhibits regular 2007. The chart identifies several discrete events that may have clock effects. Indeed, as indicated by the equidistant spikes affected the timing of Fedwire payments: changes in operating in the 1998 probability distribution in Chart 2, there seems to be a flurry of activity every half-hour on the half- hours, changes in CHIPS operations, changes in CLS Bank hour between 11:30 and 16:30. In contrast, these clock operations, and changes in the Federal Reserve’s Payments effects are not as discernible in the 2006 distribution. An System Risk Policy regarding GSEs. Chart 6 provides a slightly analysis of the distribution of value settled for each year different perspective on how the timing of payments evolved between 1998 and 2006 indicates that the clock effects over time. Five points in particular are worth noting: gradually dissipate until they virtually disappear in 2002. 1. The deciles exhibit different trends. The first two deciles We have not been able to identify what causes these clock show a negative trend while the deciles between effects and why they have faded away over time. In 40 percent and 90 percent indicate a slightly positive particular, it is unclear whether the effects are attributable trend. These results are consistent with the dual shift we to technological factors, the behavior of Fedwire identified earlier when comparing the value of payments participants, or institutional constraints. in 2006 and 1998. Indeed, we found that in 2006 the value 5. The 2006 distribution exhibits a higher amount of settled moved toward earlier payments at the beginning activity at precisely 08:00 and 08:30. An analysis of the of the day and toward later payments later in the day.
Chart 6 Deciles of Fedwire Funds Value Time Distribution
January 18, 2000: January 22, 2001: December 8, 1997: CHIPS CHIPS September 10, 2002: May 17, 2004: July 1, 2006: Opening at 00:30 Closing at 17:00 Intraday Finality CLS Bank Opens Opening at 21:00 GSE Policy
Closing 18:30 100% 18:00 90 17:00 80 70 60 16:00 50 15:00 40 30 14:00 13:00 20
12:00 10 11:00 10:00 1997 98 99 00 01 02 03 04 05 06
Sources: Federal Reserve Bank of New York; authors’ calculations. Notes: A twenty-one-day centered moving average is used. Values exclude payments associated with CHIPS, CLS Bank, DTC, and principal and interest payment funding. GSE is government-sponsored enterprise.
90 Changes in the Timing Distribution 2. The first four deciles are not grouped together as much and compared the results of several alternative specifications to as the five deciles between 50 percent and 90 percent are. ensure that our results are robust. This observation is consistent with the shape of the Our sample consists of daily observations for virtually every payment value time distribution in Chart 2, which is business day between March 1998 and November 2006.6 In a highly concentrated toward the end of the day. given regression, the dependent variable is defined as the time 3. After 2004, the 100th percentile rarely exceeds the 18:30 at which the corresponding percentile of value settled on a closing time, indicating that extensions of Fedwire specific day, which we measure in the number of seconds since operating hours occur less frequently. the day’s Fedwire opening. The same set of explanatory 4. The deciles exhibit various peaks and valleys. The reasons for some of these peaks and valleys are clear, such as September 11, 2001, which led to later payments. Others Our regression analysis identifies the are not as obvious. factors that affected the distribution, 5. The events documented in Chart 6 do not have a clear or at least part of the distribution, effect on the evolution of the times at which the various deciles of payment value settle. In particular, it is difficult of payment values. to conclude unambiguously from the chart whether trends in the percentiles of value can be imputed directly to any one of these events. In the next section, we conduct variables is used in each of the 100 regressions. A formal a regression analysis to disentangle the effects of various definition of these variables as well as their sources can be factors on the timing of payments. found in Appendix A. Drawing on the literature we reviewed earlier, we include a number of potentially relevant variables in our analysis, which we organize into five categories: value and volume, Federal Reserve policies and operations, settlement 4. Regression Analysis system activities, other control variables, and calendar effects. Summary statistics for the independent variables are presented in Table 2. 4.1 Model and Data
Our regression analysis identifies the factors that affected the Value and Volume distribution, or at least part of the distribution, of payment The value as well as the number, or volume, of payments values. After experimenting with different specifications, we transferred over Fedwire may play a role in determining when settled upon an easily interpretable yet robust model consisting Fedwire participants submit payments. To account for these of 100 linear regressions, each estimated separately for a given effects, we disaggregate the nonsettlement daily value of percentile of value.4 To address possible serial correlation and Fedwire funds into four mutually exclusive groups. More heteroskedasticity problems, we relied on the approach specifically, we differentiate: 1) the total value transferred by developed by Newey and West (1987) to correct the estimated banks on behalf of their customers, 2) the total value of standard errors.5 Finally, we conducted various diagnostic tests deliveries of federal funds purchases and sales, 3) the total value of federal funds returns (of the prior day’s deliveries), and 4 We recognize that there may be better specifications as well as more efficient inference techniques for analyzing the Fedwire value time distribution. In 4) the value of all other interbank transfers, thereby consisting particular, since we do not estimate the joint distribution of all percentiles, we of payments not included in the groups made on behalf of are not able to compare statistically point estimates across neighboring customers or as part of a federal funds purchase, sale, or return. percentiles. Instead, we contrast only how a variable of interest affects different parts of the distribution, such as the low percentiles (corresponding to the All else equal, and controlling in particular for the number of morning) and the high percentiles (corresponding to the late afternoon). Fedwire payments, we observe that a higher value of transfers Observe, however, that: 1) our specification does not imply that the times at should result in a higher demand for daylight credit, which which the percentiles of value settle are independent and 2) if we can assume that our system of regression equations has the structure of a Seemingly would lead to a higher shadow cost of liquidity. As a result, Unrelated Regressions (SURE) model, then there is no loss of efficiency in one may anticipate that higher values of payments lead to later estimating the regressions separately rather than jointly by GLS, since the settlement of payments. explanatory variables are the same in each percentile regression. 5 Note that we also estimated the model with lagged (up to ten lags) dependent 6 We are missing data for the following dates: April 1, 1997; December 22-24, variables. The results remain virtually unchanged and the differences are 1997; March 1-3, 1999; April 20-22, 1999; October 14-15, 1999; October 18, strongly insignificant. 1999; and November 9, 1999.
FRBNY Economic Policy Review / September 2008 91 Table 2 Summary Statistics of Independent Variables
Variable Mean Median Standard Deviation Minimum Maximum
Target federal funds rate (percent) 3.61 4.00 1.88 1.00 6.50 Operating hour extension (minutes) 00:04:05 00:00:00 00:19:06 00:00:00 05:16:00 Interbank payment value (billions of dollars) 487.08 476.19 75.40 148.79 865.82 Customer payment value (billions of dollars) 610.25 585.75 144.39 152.27 1334.25 Federal funds deliveries (billions of dollars) 250.83 257.65 65.70 0.92 472.64 Federal funds returns (billions of dollars) 250.09 257.84 64.21 0.92 432.61 Payments greater than or equal to $10 million 0.908 0.909 0.011 0.853 0.934 Number of payments (thousands) 465.237 453.817 79.101 186.895 904.726 Federal funds deviation -0.01 0.00 0.13 -1.56 1.81 HHI of Fedwire value 529.8 516.5 107.0 220.2 795.3
Sources: Federal Reserve Bank of New York; authors’ calculations. Notes: Payments greater than or equal to $10 million is defined as the fraction of daily value from payments greater than or equal to $10 million, excluding CHIPS, CLS Bank, DTC, and principal and interest funding payments. HHI is the Herfindahl-Hirschmann Index, a measure of the concentration of payment activity among banks.
As the size of payments varies greatly, we also include in Risk Policy. These upper limits on the amount of daylight the regressions the share of daily value consisting of individual overdrafts that can be extended to a bank participant are called payments in excess of $10 million in value. We chose net debit caps. We include as a dummy variable the date of the $10 million as a threshold because it is quite close to the top liberalization of net debit caps that occurred on February 21, 1 percent of individual payments when ranked by value. We 2002, which allowed foreign banking organizations to modify hypothesize that a large proportion of high-value payments the net debit cap calculation for U.S. branches and agencies of would cause a greater need for daylight credit, resulting in foreign banks (we call that variable the Foreign Capital higher shadow costs of liquidity and later payments. Equivalency Policy). That change resulted in a one-time 7 Finally, we control for the number of payments. After we increase of approximately 10 percent in the aggregate net debit hold the values of payments constant, an increase in volume caps of Fedwire participants. We argue that this change in results in smaller individual payments on average. Therefore, policy lowered the costs of liquidity for these banks and should one may expect that higher volume would be associated with result in earlier settlement of payments. Because the policy faster Fedwire settlement, as smaller individual payments are change was applied primarily to foreign banks—many of which less likely to require drawing in daylight credit. An alternative participate in CHIPS and CLS Bank, which have early-in-the- hypothesis is that a higher volume of transfers places higher day activity—we expect this faster settlement to affect mainly operational demands on banks to check the credit lines of the lowest percentiles of Fedwire activity. customers and other processes associated with submission Another change in the Payments System Risk Policy, which of payments to Fedwire, and it could therefore result in later occurred on July 20, 2006, restricted GSEs and certain settlement. international organizations from incurring daylight overdrafts. In general, this change represents a restriction of access to daylight credit, and we would expect it to correspond to an Federal Reserve Policies and Operations increase in liquidity costs. Thus, we hypothesize that this The Federal Reserve Banks offer many Fedwire participants change, which we capture with a dummy variable, would result access to daylight overdrafts for a fee and subject to upper in slower payment settlement. Not all GSEs are exactly limits, as described in the Federal Reserve’s Payments System comparable, however. In particular, a distinctive feature of Fannie Mae and Freddie Mac is their payment of principal 7 Note that although the value and volume variables drift during the sample and interest on the 15th and the 25th of each month (or on period, we find no evidence suggesting that they may be nonstationary. More dates close to the 15th and 25th when they fall on weekends or specifically, after conducting a series of Dickey-Fuller tests (with ten lags) and Phillips-Perron unit-root tests, we rejected the nonstationary hypothesis at the banking holidays). We therefore include an additional dummy 1 percent significance level. variable to measure whether the change in policy affects the
92 Changes in the Timing Distribution timing of Fedwire payments differently on these specific dates. payments (we discuss these points in more detail below). We We do not have an unambiguous prediction about the effect of conjecture that the time of these settlements will positively this policy change on the timing distribution. On the one hand, influence the timing of Fedwire payments: as their times move, the change may generate more delays, as they represent a so will the timing of Fedwire payments. restriction on access to liquidity; on the other hand, it may In addition to time variables associated with CHIPS, DTC, accelerate Fedwire payments, because GSEs must have funds and CLS Bank, we include the value of U.S. dollar settlements delivered to them prior to releasing their principal and interest conducted each day through CLS Bank, the values of the initial payments. and final prefunding values in CHIPS, and the net-net We include a dummy variable for all dates after the Federal Reserve extended, in May 2004, the opening hours of Fedwire to 21:30. We contend that this change may increase the [Our analysis includes] variables submission of early-morning payments to Fedwire. measuring the times at which CHIPS Occasionally, the Federal Reserve Banks decide to extend the hours of Fedwire operation because of significant operational and DTC conduct their late-afternoon problems of a participant or the system. As such, we include settlements . . . . These times can also a variable that measures the duration of the extension in act as focal points for the settlement minutes. We hypothesize that extensions may increase settlement risk and uncertainty and are associated with later of other Fedwire payments. payments—especially for the final few percentiles of payment value—and therefore they slow the settlement of the later percentiles of funds transfers. Finally, we include the target fed settlement values in DTC. In our view, an increase in values funds rate as a variable to control for any effect that monetary settled through the settlement systems would increase demand policy decisions might have on the timing of payments. for daylight credit, increasing its shadow cost, and result in later settlement times.
Settlement System Activities Every day, most financial institutions are active simultaneously Other Control Variables in a number of markets and payments systems—in particular, We include both a constant and a time trend in our regression. in the settlement systems CLS Bank, CHIPS, and DTC. (See The time trend is meant to control for trends, such as Appendix B for a description of these systems.) This activity technological change, other than those captured by other may affect the liquidity available to these participants at a given covariates (for example, the volume and values of payments). time during the day, which in turn may influence the time at In addition, we include the Herfindahl-Hirschmann Index which Fedwire participants decide to submit payments. To (HHI) of payment market shares, a measure of the control for the influence of settlement system activity on the concentration of payment activity among banks. We expect timing of Fedwire payments, we include a dummy variable for this variable to control for industry mergers and other changes the dates after which CLS Bank began operation. Because CLS in the pattern of payments between banks. As discussed earlier, Bank operates early in the day in the United States, we reason industrial structure can affect payment timing in a number of that it may quicken the settlement of Fedwire payments ways not fully examined in the literature. A more concentrated submitted in the morning. industrial structure might be able to coordinate payments We also include variables measuring the times at which more easily, but at either an earlier or later time of day. A more 8 CHIPS and DTC conduct their late-afternoon settlements. concentrated structure could result in more payment value We hypothesize that the times of these settlement systems are being transferred by the larger bank that is more likely to have associated with decreased uncertainty and with rapid exceeded its deductible portion of the overdraft fee schedule, redistribution of balances in various banks’ accounts after these and therefore is more likely to economize liquidity actively. settlements are complete. Because of these effects, these times As a result, a greater concentration in industrial structure in can also act as focal points for the settlement of other Fedwire the payment market could lead to later settlement. Finally, we include the interest rate spread between the effective federal 8 CLS Bank settles at multiple times in the early morning; because CLS Bank operates so early and its settlement time is so diffuse and therefore difficult to funds rate and the target federal funds rate. We reason that characterize, we do not include a settlement time variable for it. when this spread is high, the net demand for end-of-day
FRBNY Economic Policy Review / September 2008 93 balances is relatively high, which we would expect to be hours after the opening of Fedwire, is only the 15th percentile of associated with later payment timing. value time. By comparison, the hour between 16:00 and 17:00 includes more than twenty percentiles. Each chart corresponds to an explanatory variable; for instance, Chart 7.1 corresponds to interbank payment value. Each chart plots twenty points Calendar Effects indicating the point estimate of the coefficient for that variable There are many predictable differences in payment activity in the corresponding linear regression. As indicated in the chart across the days of the week and over the year. For example, notes, the color of the point identifies the level of statistical Mondays predictably have higher volume than Fridays on significance of the point estimate. On each side of a point average; days at the beginning of the month are likewise estimate, we add a band representing the 95 percent confidence expected to be high value. To control for these effects on the interval for this point estimate. Finally, we plot in Chart 12 the timing of Fedwire payments, we include a number of dummy adjusted R2 for each regression. variables for various calendar effects. Included are dummies In terms of interpretation, a parameter significantly for the days of the week, the days preceding and following greater (lower) than zero in a regression for a given percentile holidays, the first of the month, the last day of the quarter, the indicates that the marginal effect of the corresponding last day of the year and, separately, the last five days of the year. explanatory variable delays (accelerates) the time at which In addition, we include dummies for days on which the New that percentile settles. York Stock Exchange is closed and, separately, days on which it In these charts, the results of multiple percentiles are shown closes early. Dummies are also included for each of the final five on the same scale. This gives a full sense of each variable’s effect days of the reserve maintenance period (the day-of-week on the timing of payments across the entire day. However, as dummy captures both the effect of the day of the week and the we mentioned, comparisons between percentiles could be effect of the first five days of the reserve maintenance period); misleading and may overstate the economic effect of variables we include that variable in case the reserve maintenance period in the middle percentiles. Delaying a payment by the same influences payment activity. Finally, we include a separate amount of time becomes more costly as the end of the day dummy for the two-week period including and following approaches, when there is less time left to settle all remaining September 11, 2001. During that period, Fedwire operating payments. For example, five minutes of delay at 17:30, when hours were regularly extended and were expected to experience there is only one hour of the business day left, can be later activity than normal. considered a larger economic effect than five minutes Before analyzing our regression results, we emphasize one of delay at 12:00. point. The variables described above may be expected to affect some parts, but not necessarily all, of the timing distribution. For instance, CLS Bank value may be considered likely to affect the payment distribution early in the day, but it may not Value and Volume necessarily have a lasting effect on the final percentiles of the The parameters associated with the value of interbank distribution. Conversely, CHIPS final payout value may be payments settled over Fedwire are significant and greater than considered likely to affect only the upper tail of the value zero for virtually all percentiles (Chart 7.1). In other words, it distribution. appears that more interbank transfers over Fedwire tend to slow down the settlement of payments generally throughout the day. This result is consistent with the argument that banks have an incentive to delay their interbank payments, which are Results of high average value and may incur little delay cost because no To streamline our analysis, we present our estimation results customer may be demanding early settlement because of the graphically in Charts 7-11. We start by providing information cost of daylight credit. about the interpretation of each chart. The x-axes of each chart In contrast, the parameters corresponding to the total value display two scales. The bottom scale represents the percentile transferred by banks on behalf of their customers over Fedwire of value, or equivalently the regression number, moving from are negative and significant for all percentiles below 85 percent, 1 to 100. The top scale represents the average time in 2006 at that is, for payments submitted before 17:45 on average in 2006 which the corresponding percentile settled during the day. (Chart 7.2). Fedwire payments therefore seem to settle earlier, The afternoon peak of the Fedwire value time distribution is when the value of transactions transferred by banks’ customers evident when comparing these two scales. 12:00, which is fifteen is high. This result may be explained by the fact that, compared
94 Changes in the Timing Distribution Chart 7 Regressions of Fedwire Funds Value Time Percentiles Value and Volume
1: Interbank Payment Value 2: Customer Payment Value 3: Federal Funds Deliveries 12 13 14 15 16 17 18 12 13 14 15 16 17 18 12 13 14 15 16 17 18 1.00 0.10 0
0.05 0.05 -0.05 0
0 0.10 -0.05 020406080 100 020406080 100 020406080 100
5: Payments Greater than or Equal to 4: Federal Funds Returns $10 Million 6: Number of Payments 12 13 14 15 16 17 18 12 13 14 15 16 17 18 12 13 14 15 16 17 18 0.15 1,500 0.15 0.10 1,000 0.10 0.05 500 0.05 0 0 -0.05 0 020406080 100 020406080 100 020406080 100 Percentile Percentile Percentile
Sources: Federal Reserve Bank of New York; authors’ calculations. Notes: The upper x-axis displays the mean 2006 time for selected percentiles. The y-axis displays the value of the coefficient. The band represents the 95 percent confidence interval for the point estimates. We use independent ordinary least squares regressions with Newey-West standard errors (maximum lag = 10) for the 2nd to 99th percentiles of value. The color of the point indicates the significance of the coefficient: blue = 1 percent, light blue = 5 percent, dark gray = insignificant. There are 2,200 observations for each regression.
with interbank transfers, banks face a higher delay cost when federal funds purchases and sales also capture the demand for acting on customers’ requests for payments. In particular, overnight credit. This would therefore explain why higher banks may be asked by their customers to execute their federal funds deliveries are associated with delayed Fedwire transfers by a certain time. payments throughout the day. Likewise, it is possible that The results suggest that both forms of federal funds activity banks expecting a return of federal funds may tend to delay tend to delay the timing of Fedwire payments. Indeed, the their Fedwire payments in the morning until their accounts significant parameters in Charts 7.3 and 7.4 are systematically have been credited. We now turn to the effect of the number of Fedwire payments transferred in a day. Virtually all parameters in All else equal, an increase in the number Chart 7.6 are positive and significant for percentiles up to 80 percent, or equivalently for payments transferred before of Fedwire transfers results in delayed 17:45 on average in 2006. In other words, all else equal, an payments for most of the day. increase in the number of Fedwire transfers results in delayed payments for most of the day. This result seems to contradict our hypothesis that a greater number of transfers may expedite greater than zero. Observe, however, that federal funds returns Fedwire payments, as it implies lower average size payments appear to have a slightly larger effect earlier in the day, while the once we control for the total value transferred. Instead, the effect of federal funds deliveries persists throughout the day. result possibly points toward greater operational costs, These results may be considered surprising since both types whereby banks must delay payments because it takes more of activities tend to occur later in the day. We conjecture that time to process a greater number of payments.
FRBNY Economic Policy Review / September 2008 95 The size of individual payments, however, is not completely parameters for percentiles above 35 percent are significant and neutral. Indeed, we find that the fraction of individual payments greater than zero. This result is therefore consistent with our exceeding $10 million affects the timing of payments for hypothesis that the removal of access to intraday credit by GSEs percentiles up to 65 percent (Chart 7.5). In other words, large and international organizations resulted in a shift toward later individual transfers delay payments submitted before 16:45 on Fedwire payments. It is also consistent with the observation average in 2006. This result does not unambiguously support our that the Federal Home Loan Banks decided to delay settlement view that large individual payments lead to delayed settlement of their principal and interest payments from 08:30 to because they increase the likelihood of daylight overdrafts. We approximately 14:00 after the implementation of the policy argue that our finding of no such delays after 16:45 may be (but the delay effects persist throughout most of the remainder explained by the fact that opportunities for multilateral netting of the day). are greater at the peak of Fedwire activity. As a result, banks may As Fannie Mae and Freddie Mac are somewhat distinct from be less likely to delay large individual payments at the end of the other GSEs, we also test whether the change in the Payments day, as the risk of daylight overdraft decreases. System Risk Policy had a specific effect on the 15th and 25th of the month, dates on which these two institutions make their principal and interest payments. Controlling for these specific Federal Reserve Policies and Operations dates over our entire sample period (see our discussion below) The July 1, 2006, modifications to the Federal Reserve’s as well as for the policy change for all days following its Payments System Risk Policy with regard to GSEs seem to have implementation, we find that the timing of Fedwire payments contributed to the delays in payments submitted after 15:00 shifted to earlier in the morning, but remained unchanged in on average in 2006 (Chart 8.2). Indeed, most estimated the afternoon. Indeed, only the percentiles below 20 percent
Chart 8 Regressions of Fedwire Funds Value Time Percentiles Federal Reserve Policies and Operations
1: Foreign Capital Equivalency Policy 2: GSE Credit Policy 3: MBS P&I Day * GSE Policy 12 13 14 15 16 17 18 12 13 14 15 16 17 18 12 13 14 15 16 17 18 -10 -10 0
0 0 -10
-10 -10 -20
-20 -20 -30 020406080 100 020406080 100 020406080 100
4: Fedwire Opens at 21:00 5: Operating Hour Extension 6: Federal Funds Target Rate 12 13 14 15 16 17 18 12 13 14 15 16 17 18 12 13 14 15 16 17 18 20 0.8 2
0.6 0 10 0.4 -2 0 0.2 -4
-10 0 -6 020406080 100 020406080 100 020406080 100 Percentile Percentile Percentile
Sources: Federal Reserve Bank of New York; authors’ calculations. Notes: The upper x-axis displays the mean 2006 time for selected percentiles. The y-axis displays the value of the coefficient. The band represents the 95 percent confidence interval for the point estimates. We use independent ordinary least squares regressions with Newey-West standard errors (maximum lag = 10) for the 2nd to 99th percentiles of value. The color of the point indicates the significance of the coefficient: blue = 1 percent, light blue = 5 percent, dark gray = insignificant. There are 2,200 observations for each regression. GSE is government-sponsored enterprise; MBS is mortgage-backed securities; P&I is principal and interest.
96 Changes in the Timing Distribution are significant and negative (Chart 8.3). This result may be Bank (Chart 9.8) or the variable capturing the value of pay- explained by a combination of two factors. First, compared ments exchanged over CLS Bank (Chart 9.7) are insignificant. with other GSEs, Fannie Mae and Freddie Mac did not delay Therefore, the conjecture that the creation of CLS Bank may markedly their payments of principal and interest after July 1, have helped speed up other Fedwire payments because it settles 2006. Second, after the policy change, these two GSEs found early in the day turned out to be unfounded. ways to have funds delivered to them earlier in the morning In contrast, the other two settlement systems—CHIPS and than they did prior to the change, in order to avoid having to DTC—appear to play major roles in shaping the value time draw on daylight overdrafts. distribution of Fedwire funds transfers, especially toward the In another significant change to the Payments System Risk end of the day. In particular, we identify significant delays in Policy, the February 21, 2002, Foreign Capital Equivalency Fedwire payments submitted late in the afternoon (between Policy change increased the net debit cap of foreign banking 16:15 and 17:15 on average in 2006) on days DTC settles later. organizations. We hypothesize that because foreign banking organizations conduct a larger percentage of their payments in the early Fedwire operating hours (from 21:00 to 08:00), which CHIPS and DTC appear to play major overlap with the operating hours of European and Asian roles in shaping the value time distribution markets, this change would affect the low percentiles of the of Fedwire funds transfers, especially Fedwire value time distribution. Chart 8.1 shows that the estimated parameters for percentiles below 20 percent are toward the end of the day. significant and negative, indicating that the change in Federal Reserve policies accelerated the submission of Fedwire payments in the morning. This result is therefore consistent Indeed, the point estimates in Chart 9.6 are significantly greater with our prediction that the increase in net debit caps benefited than zero in the regressions conducted for the percentiles mostly foreign banks submitting payments early in the 55 percent to 85 percent. As we discuss in the next section, this morning, partly because of their active participation in CHIPS result is particularly relevant, as DTC settlement typically and CLS Bank. occurs near the time of highest Fedwire activity. We also find that the parameters associated with the Likewise, we find that the time of CHIPS settlement plays duration of occasional extensions of Fedwire opening hours a significant role in explaining the upper tail of the value time are significant and positive for most percentiles throughout distribution of Fedwire payments. Indeed, the estimations the day. Observe also that the magnitude of the effect is reported in Chart 9.1 suggest a strong positive effect, highly significantly larger as the official closing time nears (that is, concentrated around the 75th percentile, which is very close to after 17:30). In other words, and in line with intuition, the the time at which the highest peak of Fedwire activity occurs in duration of an extension is in general positively correlated 2006 (Chart 2). As we explore in greater detail in the next with delays in the timing of payments, and it is particularly section, the emergence of the after-17:00 peak in Fedwire value 9 powerful for explaining delays at the end of the day. transferred coincides with the change in CHIPS settlement Finally, and somewhat surprisingly, our regression results time. In other words, the end-of-day shift in the timing of in Chart 8.4 suggest that payments submitted in the morning payments toward a later time may be traced in large part to were settled significantly later, not sooner, after Fedwire the change in the timing of CHIPS settlement. Our regression extended its opening hours from 00:30 to 21:00. results suggest that the January 2000 change in the timing of CHIPS settlement led to later settlement of the 65th-95th percentiles of Fedwire value. Settlement System Activities In contrast with the effects of the change in CHIPS We find no evidence supporting the hypothesis that the settlement time, the change in the CHIPS settlement opening of CLS Bank had an effect on the timing of payments mechanism (to provide intraday finality of payments made via submitted through Fedwire (again, recall that payments to and CHIPS) quickened the settlement of Fedwire payments from CLS Bank itself are removed from our measures). Indeed, throughout most of the day. Most of the estimated parameters all but a small number of estimated parameters associated with in Chart 9.2 are significant and negative. This result may point either the dummy variable capturing the opening date of CLS to a consequence that the new CHIPS settlement mechanism has on customers: It may enable banks to credit their customers 9 We find no indication that the occasional extensions of Fedwire operations may be endogenous. In other words, this variable may be considered as for payments made on CHIPS earlier in the day than was the capturing only technical failure in Fedwire operations. practice before January 22, 2001. As a result of this earlier
FRBNY Economic Policy Review / September 2008 97 Chart 9 Regressions of Fedwire Funds Value Time Percentiles Settlement Institutions
1: CHIPS Closes at 17:00 2: CHIPS Intraday Finality 3: CHIPS Final Payout Value 12 13 14 15 16 17 18 12 13 14 15 16 17 18 12 13 14 15 16 17 18
10 10 0.2 5 0 0.1 0 -10 0 -5 -10 -20 -0.1
020406080 100 020406080 100 020406080 100
4: CHIPS Extension 5: DTC Net-Net Credits 6: DTC Settlement Time 12 13 14 15 16 17 18 12 13 14 15 16 17 18 12 13 14 15 16 17 18
20 0.2 0.2 0 0.1 10 -0.2 0 0 -0.4 -0.1 020406080 100 020406080 100 020406080 100 Percentile 7: CLS Bank U.S. Dollar Value 8: CLS Bank Opens 12 13 14 15 16 17 18 12 13 14 15 16 17 18 0.5 10 0 5 -0.5 0 -1.0 -5 020406080 100 020406080 100 Percentile Percentile
Sources: Federal Reserve Bank of New York; authors’ calculations. Notes: The upper x-axis displays the mean 2006 time for selected percentiles. The y-axis displays the value of the coefficient. The band represents the 95 percent confidence interval for the point estimates. We use independent ordinary least squares regressions with Newey-West standard errors (maximum lag = 10) for the 2nd to 99th percentiles of value. The color of the point indicates the significance of the coefficient: blue = 1 percent, light blue = 5 percent, dark gray = insignificant. There are 2,200 observations for each regression.
redistribution of liquidity, banks’ customers may now be able 16:30), and 3) have no effect on payments submitted at the end to submit other payments over Fedwire earlier. of the day (after 16:30). We have no explanation for the first Less clear is the influence of the variables capturing CHIPS result. The second result may be explained by the fact that, and DTC values on the timing of Fedwire payments. For as the level of activity on DTC increases throughout the day, instance, we find that large CHIPS final payouts slow Fedwire liquidity available to other banks to make Fedwire payments payments at the end of the day (Chart 9.3). This result is not is removed. Finally, the third result suggests that at the end consistent with the hypothesis that, by releasing funds after of the day, the timing of Fedwire payments is affected only by it settles, CHIPS may accelerate payments made through the time at which DTC settles, not by the value of DTC net- Fedwire. Instead, it could suggest that CHIPS participants that net credits. have to make payments to settle their positions may experience a temporary liquidity squeeze that leads them to delay their Fedwire payments. Alternatively, it may reflect some greater uncertainties in banks’ positions on days of high CHIPS Other Control Variables settlements that cause increased delays of Fedwire payments. We find that a higher degree of industry concentration, as The effect of DTC net-net credits is complicated. Chart 9.5 measured by the HHI, slows the transfer of Fedwire payments shows that larger DTC net-net credits appear to: 1) expedite submitted after 12:00 up until the time of highest Fedwire Fedwire payments submitted in the morning (before 13:30), activity. Indeed, most of the estimated parameters below the 2) slow mid-afternoon Fedwire payments (between 15:30 and 75th percentile in Chart 10.2 are positive and significant. One
98 Changes in the Timing Distribution Chart 10 Regressions of Fedwire Funds Value Time Percentiles Other Control Variables
1: Fed Funds Deviation 2: HHI of Fedwire Value 12 13 14 15 16 17 18 12 13 14 15 16 17 18
10 0.10
0 0.05
-10 0
-0.05 -20
020406080 100 020406080 100
3: Time Trend 4: Constant 12 13 14 15 16 17 18 12 13 14 15 16 17 18
0 1,000 -0.01 500 -0.02
-0.03 0
-0.04 -500 020406080 100 020406080 100 Percentile Percentile
Sources: Federal Reserve Bank of New York; authors’ calculations. Notes: The upper x-axis displays the mean 2006 time for selected percentiles. The y-axis displays the value of the coefficient. The band represents the 95 percent confidence interval for the point estimates. We use independent ordinary least squares regressions with Newey-West standard errors (maximum lag = 10) for the 2nd to 99th percentiles of value. The color of the point indicates the significance of the coefficient: blue = 1 percent, light blue = 5 percent, dark gray = insignificant. There are 2,200 observations for each regression. HHI is the Herfindahl-Hirschmann Index, a measure of the concentration of payment activity among banks.
hypothesis is that an increase in industry concentration helps period. This shows that payment activity has moved from improve coordination and thereby facilitates the transfer of relatively small banks (market shares ranked below 50th), payments around a single point in time. Our finding that which face low marginal overdraft costs, to the largest banks increased concentration in payment market shares tends to (top four), which regularly face a higher positive fee for the slow the settlement of payments might still be consistent use of their marginal daylight overdrafts. with increased coordination, but it reflects the fact that The federal funds rate deviation shifts most of the distri- coordination occurs around a later time of day, possibly bution of Fedwire payments earlier (Chart 10.1). This result at the peak of Fedwire activity. does not support the hypothesis that a positive deviation An alternative hypothesis is that the increased concen- reflects a higher than anticipated demand for intraday tration results in greater economization of liquidity by banks. liquidity by Fedwire participants and therefore would The largest banks are more likely to pay positive overdraft be associated with later Fedwire payments. Instead, we fees and therefore face a positive marginal cost of liquidity. conjecture that the effect of federal funds deviations could As payment shares move from banks with a zero marginal cost be explained by Fedwire participants having an incentive to of liquidity to those with a positive marginal cost, we would purchase federal funds early if they are trading at a higher expect settlement to occur later in the day. The nature of the than anticipated price. increase in concentration is consistent with this hypothesis; the Finally, the time trend is found to be significantly lower than payment value market share of the top four banks has increased zero for most percentiles during the day (Chart 10.3). Before by 13 percentage points, while the share of banks ranked 5th we interpret this result, recall that in the descriptive analysis we through 50th has declined by 2 percentage points over the identified a dual adjustment process between 1998 and 2006,
FRBNY Economic Policy Review / September 2008 99 with a trend toward earlier payments for low percentiles and 4.2 Economic Significance a trend toward later payments for higher percentiles. The estimated time trend therefore appears to capture part of the To put our regression results into a more general perspective, first effect, but it is not consistent with the second. In other we conduct two exercises. First, we measure which variables words, the move toward later Fedwire payments at the end of can explain a later-than-normal settlement of Fedwire value the day is explained in our regressions by explanatory variables on a given day in 2006. The second exercise measures the other than the exogenous time trend. In addition, observe that approximate economic contribution of various factors in the influence of the time trend on the timing of Fedwire explaining the shift in late-day payments between 1998 and transfers provides some support for the hypothesis that 2006. To start, both exercises confine our attention to those technological improvements, such as in queuing mechanisms variables that largely explain the variation in the timing of used at various banks, may have contributed to accelerating payments submitted after 17:00. This may be considered transfers over Fedwire. particularly relevant in light of the Board of Governors of the
Calendar Effects Understanding the factors affecting We now comment briefly on some of the major calendar the submission of Fedwire payments effects. We find that compared with Thursdays, the timing at the end of the day is . . . of particular of payments on Mondays (especially in the morning) and Tuesdays is delayed (Charts 11.5 and 11.6), but for the most interest when analyzing potential part the timing is not significantly different on Wednesdays economic costs and benefits and Fridays, except for a marked end-of-week effect at the close of alternative policy options. of Fedwire on Fridays (Charts 11.7 and 11.8). Controlling for the days of the week, we find that the timing of Fedwire payments is virtually identical during the second week of the Federal Reserve System’s 2006 consultation paper, which maintenance period (Charts 11.9-11.12), except for the points out that the recent shift of the Fedwire activity peak to Monday of the second week (Chart 11.11), when payments closer to the end of the day raises some concerns in terms of settle earlier. operational problems. Understanding the factors affecting the Chart 11.1 indicates that Fedwire payments tend to settle submission of Fedwire payments at the end of the day is earlier when Fannie Mae and Freddie Mac make principal and therefore of particular interest when analyzing potential interest payments (on the 15th and 25th of the month). This economic costs and benefits of alternative policy options. effect was anticipated, because these payments are typically In our first exercise, we take out the effect of long-run issued around 08:30, and therefore they provide Fedwire growth in most continuous variables, such as volume and participants with an influx of liquidity early in the morning. value, by removing the trend of each of the continuous In addition, we find that, all else equal, Fedwire payments settle earlier: 1) on days when the New York Stock Exchange is covariates (those that vary in number or in value over time). either closed or closes early (Charts 11.2 and 11.3), 2) on the Next, we measure a small day-to-day variation in the level of days preceding and following a holiday (Charts 11.14 and the variable—namely, a one standard deviation of this adjusted 11.15), 3) on the last days of each quarter (Chart 11.17), and variable. We also consider a typical day by setting all calendar 4) on the last five days of the year (Chart 11.18). In contrast, effects equal to zero and without an extension of Fedwire Fedwire payments tended to be submitted later on the first operation. Finally, we ignore the effect of past specific events, of the month (Chart 11.16) and during the week following such as the May 17, 2004, extension of Fedwire operating hours September 11, 2001 (Chart 11.4). and the creation of CLS Bank, as these events are not expected Finally, observe that the adjusted R2s are generally high to repeat in the variation of activity on Fedwire from one day to (Chart 12), indicating that our regression models are able to the next. We then measure the economic significance of all of capture a large part of the daily variations in the percentiles of our variables in explaining the timing of payments made on Fedwire value. Note also that the adjusted R2s tend to be closer Fedwire after 17:00 by multiplying the estimated coefficient by to 1 for higher percentiles. This result is consistent with the fact the one-standard-deviation change in the variable. that low percentiles, corresponding to payments settled in the Surprisingly, we find that only three variables appear to play morning, are in general more volatile from one day to the next. a significant economic role in explaining the timing of Fedwire
100 Changes in the Timing Distribution Chart 11 Regressions of the Fedwire Funds Value Time Percentiles Calendar Effects
1: MBS P&I Day 2: New York Stock Exchange Closed 3: New York Stock Exchange Closed Early 12 13 14 15 16 17 18 12 13 14 15 16 17 18 12 13 14 15 16 17 18 5 0 0 0 -10 -5 -50 -20 -10 -30 -15 -100 020406080 100 020406080 100 020406080 100
4: September 11-18, 2001 5: Monday 6: Tuesday 12 13 14 15 16 17 18 12 13 14 15 16 17 18 12 13 14 15 16 17 18
200 15 10 10 5 100 5 0 0 0 -5 020406080 100 020406080 100 020406080 100
7: Wednesday 8: Friday 9: Maintenance Day 6 12 13 14 15 16 17 18 12 13 14 15 16 17 18 12 13 14 15 16 17 18 6 5 4 4 2 2 0 0 0 -2 -5 -2 020406080 100 020406080 100 0204060 80 100
10: Maintenance Day 7 11: Maintenance Day 8 12: Maintenance Day 9 12 13 14 15 16 17 18 12 13 14 15 16 17 18 12 13 14 15 16 17 18 2 2 2 0 0 0 -2 -2 -4 -2 -4 -6 -4 020406080 100 020406080 100 020406080 100
13: Maintenance Day 10 14: Day after a Holiday 15: Day before a Holiday 12 13 14 15 16 17 18 12 13 14 15 16 17 18 12 13 14 15 16 17 18 2 10 0 0 0 -10 -2 -10 -4 -20 -20 020406080 100 020406080 100 020406080 100
16: First Day of Month 17: Last Day of Quarter 18: Last Five Days of Year 12 13 14 15 16 17 18 12 13 14 15 16 17 18 12 13 14 15 16 17 18 5 10 10 0 0 0 -5 -10 -10 -10 -15 -20 -20 020406080 100 020406080 100 020406080 100 Percentile Percentile Percentile
Sources: Federal Reserve Bank of New York; authors’ calculations. Notes: The upper x-axis displays the mean 2006 time for selected percentiles. The y-axis displays the value of the coefficient. The band represents the 95 percent confidence interval for the point estimates. We use independent ordinary least squares regressions with Newey-West standard errors (maximum lag = 10) for the 2nd to 99th percentiles of value. The color of the point indicates the significance of the coefficient: blue = 1 percent, light blue = 5 percent, dark gray = insignificant. There are 2,200 observations for each regression. MBS is mortgage-backed securities; P&I is principal and interest.
FRBNY Economic Policy Review / September 2008 101 Chart 12 Adjusted R 2 of Regressions
1.00 0.90 0.80
0.70 0.60 0.50 0.40 0.30
0.20 0.10 0 0 10 20 30 40 50 60 70 80 90 100 Percentile of value time
Source: Authors’ calculations.
payments submitted after 17:00 from one day to the next. In to disentangle the respective contribution of each explanatory particular, a typical day-to-day variation in either the value of variable from the changes in the timing distribution observed interbank payments or in the number of daily Fedwire transfers during this period. Because most of the covariates varied jointly each delays payments submitted after 17:00 by roughly three between 1998 and 2006, our model cannot pin down precisely minutes on average. In addition, we find that an extension of the contribution of each explanatory variable to the shift in CHIPS operating hours delays virtually all payments submitted the timing of payments. The results presented here should after 17:00 by ten minutes on average. We note that although therefore be interpreted as orders of magnitudes rather than large in magnitude, this effect does not necessarily constitute a exact measurements. Again, we are more interested in the major risk, as CHIPS extensions are rare in practice, especially end-of-day changes and we therefore focus on the changes in since CHIPS changed its settlement mechanism (for instance, the 75th percentile, which roughly corresponds to the time of CHIPS extended its hours of operation only once between highest Fedwire activity in 2006. January 2005 and December 2006). Nevertheless, this effect Our model predicts that compared with 1998, the 75th does illustrate how significantly the operations of settlement percentile should have shifted fourteen and a half minutes later in 2006. This shift is slightly more than the thirteen minutes we actually observe in the data, but it is within two standard [Our result] suggests that very few deviations. If we consider the effect of the time trend as being variables have an influence on payments exogenous, we find the following: submitted after 17:00 on a day-to-day basis. 1. When combined, the increase in the number and value of Fedwire payments between 1998 and 2006 accounted for slightly more than 40 percent of this shift in the 75th institutions are interconnected. To summarize, this first exercise percentile toward a later time.10 indicates that although numerous factors contribute to the shift 2. The change in CHIPS closing time on January 18, 2000, if in the time of highest Fedwire activity between 16:30 and 17:00, considered an exogenous event, contributed more than their effects appear to be confined to that period and do not 30 percent by itself. This effect, however, is partially offset spill over near Fedwire’s closing time. This result should be by the modification in the CHIPS settlement mechanism, reassuring, as it suggests that very few variables have an influence which moved the 75th percentile earlier. As a result, the on payments submitted after 17:00 on a day-to-day basis. aggregate contribution of CHIPS to the later settlement Our second exercise evaluates the economically significant 10 factors that contributed to the shift in the after-17:00 value In this measure, we include all variations between 1998 and 2006 in the values of interbank payments, customer payments, and federal fund deliveries time distribution of Fedwire payments between 1998 and 2006. and returns. We also include the increase in the share of individual payments Note, however, that our regression model is not perfectly suited greater than $10 million.
102 Changes in the Timing Distribution of the 75th percentile may be estimated at around the peak in Fedwire value transfer occurred before 17:00 in 10 percent. 1998 and after 17:00 in 2006. 3. Finally, the last major contributor to the shift of the 75th To illustrate the dependency of the Fedwire value time percentile toward a later time is the higher concentration distribution on the behavior of CHIPS settlement timing, we of payment activity among banks. We find that between measure the timing of the peak of Fedwire activity as the 1998 and 2006, the increase in the HHI accounted for midpoint of the ten contiguous minutes of highest value close to 30 percent of the shift. transferred during the day. Chart 13 displays the time of the peak ten minutes of value transferred on Fedwire from 1997 through 2006. Each point represents the time at which the midpoint of the top ten contiguous minutes of Fedwire value 5. Influence of CHIPS and DTC settled on each day between 1997 and 2006. We see that prior on Settlement Time to January 18, 2000, there was a peak at approximately the same time daily (although that time varied from day to day). Its time Our regression analysis points to a few specific variables as trended downward from around 17:00 in 1997 to around 16:48 highly explanatory of the shift to later settlement of the 70th- in early January 2000. After January 18, 2000, however, a 90th percentiles of Fedwire activity. In this section, we present distinctly new pattern emerged. Peaks tended to occur at two other evidence of the influence of the settlement institutions— specific times: 16:30 and 17:11, with the most common peak at CHIPS and DTC—on the value time distribution of Fedwire 17:11. This is consistent with the observation in Chart 5 that activity. Specifically, we consider the time at which the peak in the value time distribution of Fedwire changed from a single- Fedwire value transfer occurs. Recall Chart 2, which shows that peak distribution to a dual-peak distribution in 2000.
Chart 13 Time of Peak Ten Minutes of Fedwire Funds Value
January 18, 2000: January 22, 2001: CHIPS Moves CHIPS Intraday Closing to 17:00 Finality 19:00 Coincident with: DTC and CHIPS 18:30 DTC CHIPS Neither 18:00
17:30
17:00
16:30
16:00
15:30
15:00 1997 98 99 00 01 02 03 04 05 06
Sources: Federal Reserve Bank of New York; authors’ calculations. Notes: Seven days settling after 19:00 were excluded. Values exclude payments associated with CHIPS, CLS Bank, DTC, and principal and interest payment funding.
FRBNY Economic Policy Review / September 2008 103 Chart 14 End-of-Day Settlement Times of CHIPS and DTC
January 18, 2000: January 22, 2001: CHIPS Moves CHIPS Intraday Closing to 17:00 Finality 19:00 DTC
18:30 CHIPS
18:00
17:30
17:00
16:30
16:00
15:30
15:00 1997 98 99 00 01 02 03 04 05 06
Sources: Federal Reserve Bank of New York; authors’ calculations. Note: Seven days settling after 18:30 were excluded.
The regression analysis suggests a reason for why the shift Why should Fedwire’s peak activity in value transfer occurred. Prior to January 18, 2000, CHIPS effected its final coincide so closely with the final payouts of the major settlement at approximately 16:40; on January 18, 2000 (and settlement institutions? We advance four hypotheses, which thereafter), it moved its settlement of final payouts to are not necessarily mutually exclusive: the bank liquidity approximately 17:10, while the value-weighted time of DTC’s cascade, the customer credit cascade, the resolution of settlement remained roughly constant over the period. uncertainty, and the role played by settlement times as Examination of Charts 13 and 14 suggests that prior to focal points. January 18, 2000, Fedwire’s peak of settlement activity First, we advance our bank liquidity cascade hypothesis. occurred simultaneously with the roughly coincident Consider the activities of DTC. As we discuss in Appendix B, settlement times of CHIPS and DTC. After January 18, 2000, DTC accumulates balances in its account and releases them when CHIPS moved to a later settlement time, two peaks of back to the banking system during its final payout procedures. settlement activity emerged on Fedwire. One coincided with That outflow of balances from DTC and the resulting inflow to DTC settlement time at 16:30, and the other moved more banks can trigger a cascade of payments made by the receivers closely to CHIPS settlement time after 17:00. The distinct of DTC payouts, which triggers further payments made by change in pattern, so closely matching the timing pattern of the receivers of those payments, and so on. The cascade of CHIPS and DTC, as well as the evidence from the regression payments can occur if banks are withholding payments because analysis, points to the timing of the settlement institutions’ they face internal constraints attributable to a cost of liquidity late-in-the-day settlement as being highly explanatory in the or some other limit on their willingness to submit payments timing shifts of Fedwire’s peak and late-day activity over the earlier. Beyeler et al. (2006) provide a model and simulation 1998-2006 period. of a similar process.
104 Changes in the Timing Distribution A second reason for such a release of payments is the related Finally, banks, according to the focal point hypothesis, customer credit cascade hypothesis. Not only do banks face might coordinate their payment submissions with a settlement constraints in making timely payments, but so do their institution. Banks might choose to release payments after the customers. Those customers, or the depositors of banks, also final payouts of a major settlement institution because they receive funds following payouts on a major settlement system, believe that other banks will do so at the same time. If many which provides funds to their accounts. If those depositors banks choose to release payments at the same time, each bank had been withholding payments because they could not easily has a higher likelihood of receiving payments during that peak obtain credit to send payments, the inflow of settlement system of payment activity; therefore, it is more likely that the banks payments could provide the needed funding for them to will have a lower cost of funding their outgoing payments at execute their withheld payments; the release of those depositor that time (by incurring fewer daylight overdrafts and avoiding payments could result in the release of payments by the any constraints, such as bumping up against a net debit cap). receivers of their payments, and so on. This same phenomenon could hold true for depositors— The resolution-of-uncertainty hypothesis is a third reason a hypothesis discussed in McAndrews and Rajan (2000). why banks might release payments after a major settlement While our analysis cannot clearly distinguish between the system’s final payouts. Prior to settling at a major settlement alternative hypotheses, it is instructive to consider what the institution, banks could be uncertain of the exact amount data might imply about the relative weight of the various of their payout from the system (or their obligation to the hypotheses as explanations for the changes observed after system), and there could be some uncertainty about whether January 18, 2000. Chart 15 shows the difference in the all parties in the settlement system will perform as expected. settlement times of CHIPS and DTC and the time of the top ten After the uncertainty is resolved, banks might evaluate more minutes of Fedwire value transferred. A band of five minutes precisely the effect of releasing payments, arranging for the on either side of 0 on the y-axis represents the time of purchase or sale of federal funds, and making any other settlement of the top ten minutes of Fedwire value. A circle adjustments. represents the time of DTC settlement and a triangle the time
Chart 15 Difference between Peak Ten Minutes and Settlement Times of CHIPS and DTC
January 18, 2000: January 22, 2001: CHIPS Moves CHIPS Intraday Minutes Closing to 17:00 Finality 150
DTC CHIPS 100
50
0
-50
-100
-150 1997 98 99 00 01 02 03 04 05 06
Sources: Federal Reserve Bank of New York; authors’ calculations. Note: Values exclude payments associated with CHIPS, CLS Bank, DTC, and principal and interest payment funding.
FRBNY Economic Policy Review / September 2008 105 of CHIPS settlement. When a point falls within the five-minute 6. Conclusion band on either side of 0, the settlements of Fedwire and that system coincide. The clearest pattern that Chart 15 displays is In its report on large-value payments systems, the Committee that prior to January 18, 2000, the three systems—CHIPS, on Payment and Settlement Systems noted that the timing of DTC, and Fedwire—settled roughly simultaneously; after that payment submission is a market practice of major importance date, CHIPS and DTC settlement times diverged. The time of (Bank for International Settlements 2005). It is not uncommon Fedwire peak settlement tends to bounce between the time of for stable behavioral conventions to arise around the time DTC settlement and a time immediately subsequent to CHIPS when participants submit certain types of payments for settlement. This is because, prior to January 18, 2000, it was settlement. Such conventions can arise endogenously among fairly common to have all three systems settle roughly direct participants, and non-RTGS payments can be an simultaneously, so it is difficult to distinguish among the important “exogenous” factor affecting a bank’s RTGS various hypotheses. These same patterns are visible in Chart 13. liquidity. After January 18, 2000, we see two tendencies. First, on Our examination of payment time distribution trends in the days on which the 16:30 peak is the time of the highest value Fedwire funds service from 1998 to 2006 finds that the lower transferred on Fedwire, DTC settlement often falls within the percentiles of timing have moved to earlier in the operating day. In addition, the very last percentiles transferred have also moved to earlier over the study period, as extensions of the CHIPS is the last major institution to Fedwire operating day have become more rare. We also observe that while more value settles earlier in the morning and the settle; at that point in the processing day, later percentiles of value settle later in the afternoon, the most uncertainty about bank balances is distribution of value transferred has become more peaked. expected to be resolved and banks might A greater percentage of value was transferred during the peak of activity in 2006 than in the early part of our sample. reason that it is a good time to send We considered a host of factors affecting changes in the payments after CHIPS has settled. distribution of Fedwire timing. Federal Reserve policies and operations were found to affect settlement times in notable ways. Changes to the Federal Reserve’s Payments System Risk ten minutes of highest value transferred on Fedwire. Second, Policy in 2002, which expanded net debit caps significantly, when the 17:11 peak is the time of highest value transferred on quickened Fedwire settlement in the morning and early Fedwire, it is usually (and increasingly over time) the case that afternoon, while 2006 policy changes, which lowered CHIPS settlement precedes and falls outside the highest ten extensions of daylight credit overall, tended to slow settlement minutes of Fedwire value transferred. throughout the afternoon. Both changes are consistent with These tendencies might suggest that DTC settlement kicks banks economizing on liquidity costs in their submission of off a liquidity and customer credit cascade, an observation payments. Changes in the values and volumes transferred over supported by the quickness of value transfer on Fedwire Fedwire have increased over the period, as has the proportion following DTC settlement. This observation is also consistent of large individual payments. Taken together, these changes with the fact that DTC’s account balances grow over the day explain a large share of the later settlement of payments (after and are then released with DTC settlement to other banks, 17:00) over the period. However, we find that a larger value which effectively increases the short-term supply of liquidity in of customer payments tends to quicken Fedwire settlement, the rest of the banking system. However, the fact that the peak possibly because of the higher delay costs of customer of activity follows CHIPS settlement with only some delay payments—another influence noted in the economic might suggest that the cause of the activity is more closely literature. associated with resolution of uncertainty than with a liquidity There were numerous changes in settlement institutions or credit cascade. This explanation is also consistent with the over the period. The introduction of CLS Bank operation and fact that CHIPS settlement is mainly a redistribution of the values transferred by CLS Bank appear to have had little balances among banks, rather than a net release of liquidity effect on the value time distribution on Fedwire. Changes in back to banks. The focal point hypothesis might also offer a CHIPS operations had countervailing effects, with a 2000 move good explanation for the peak that occurs after CHIPS toward later settlement time by CHIPS clearly contributing to settlement. CHIPS is the last major institution to settle; at that a later peak in Fedwire activity. The move by CHIPS to intraday point in the processing day, most uncertainty about bank finality in 2001, though, tended to speed settlement of the balances is expected to be resolved and banks might reason that 40th-80th percentiles of Fedwire value. Increased values it is a good time to send payments after CHIPS has settled. transferred by CHIPS tended to delay settlement on Fedwire
106 Changes in the Timing Distribution for the 50th-80th percentiles, while on DTC heightened values slightly more than 40 percent to the long-run change in the quickened early settlement and delayed mid-afternoon 75th percentile of the value time distribution of payments. settlement. A payment market that has grown more concen- Of this amount, the aggregate contribution of CHIPS may be trated had a significant influence on the later settlement of estimated at around 10 percent, and the increase in industry Fedwire value. These results reflect the various calendar effects concentration accounted for close to 30 percent. and activity in other financial markets included in our The clear interdependence between payment timing on regression analysis. Fedwire and CHIPS reinforces the points made in the A major contributor to the later settlement of Fedwire Committee on Payment and Settlement Systems’ report payments was the change in CHIPS settlement time from on large-value payments systems (Bank for International roughly 16:45 to 17:10 on January 18, 2000. The time of Settlements 2005). The role of settlement institutions that the midpoint of the highest ten minutes of Fedwire value utilize Fedwire for pay-ins and payouts is a major factor transferred moved in a remarkably coincident fashion from determining system activity. This article has reviewed a 16:48 to 17:11 on that date. This pattern persisted, with the number of hypotheses on the possible channels through which Fedwire value time distribution displaying two main peaks: settlement systems affect Fedwire activity. Further research on one remaining at 16:30, nearly coincident with DTC settle- these channels would indeed provide a better understanding ment, and one at 17:11, shortly after the time of CHIPS final of the factors that affect the timing distribution of payments payouts. Over time, these new peaks of activity have in large-value systems. been stable. Also deserving of further research is the effect of increased Our results also suggest that changes in the value of industry concentration on the timing of payments, a topic that interbank payments and in the number of daily Fedwire has not been explored in the theoretical literature. Further- transfers can explain most of the daily variation in the time more, researchers could benefit from conducting similar of value transferred on Fedwire after 17:00. The rare case of studies of other payments systems to ascertain the effects of an extension of CHIPS operating hours delays virtually all daylight credit policies, system operations, settlement systems, payments submitted after 17:00, a clear illustration of the industrial structure, and other determinants of payment interdependence between Fedwire and CHIPS. timing. Their results could shed light on the robustness In addition, we estimate that increases in the number and of our results. value of Fedwire payments between 1998 and 2006 contributed
FRBNY Economic Policy Review / September 2008 107 Appendix A: Data
Our data source is Federal Reserve Bank of New York records securities (MBSs). These are generally the largest interest and of every Fedwire funds service transaction. Unless otherwise redemption payment days of the month. stated, data are used to construct the variables below associated with Fedwire funds activity. We have data on all Fedwire funds MBS P&I day, post-GSE policy is a binary variable equal to 1 transfers between April 1997 and December 2006, except for on the 15th and 25th of the month, or the first business day April 1, 1997; December 22-24, 1997; March 1-3, 1999; thereafter, after the change in GSE credit policy on July 1, 2006. April 20-22, 1999; October 14-15, 1999; October 18, 1999; and November 9, 1999. Fedwire opens at 21:00 is a binary variable equal to 1 for all days on or after May 17, 2005. On that date, the Federal Reserve extended the operating hours of the Fedwire funds service from 18 hours to 21.5 hours by moving the opening time from 00:30 to 21:00 (Board of Governors of the Federal Reserve System Variables 2003).
ith percentile of value time is the time at which i percent of Operating hour extension is the number of minutes that the the total daily value has settled. We exclude payments to Fedwire funds service remains open after 18:30. The Federal or from CHIPS, CLS Bank, and DTC. We also exclude Reserve will occasionally extend Fedwire’s operating hours at payments associated with interest and redemption payments the request of a participant having operational difficulties or of government-sponsored enterprises and international if the system is experiencing operational problems (Bank for institutions after the Federal Reserve’s Payments System Risk International Settlements 2005). Policy change on July 1, 2006. These payments related to P&I (principal and interest) are Fedwire funds payments between Fed funds target rate — Source:
GSE credit policy is a binary variable equal to 1 on and after Fed funds deliveries is the total value of new fed funds loans. July 1, 2006. The Federal Reserve changed its Payments System These loans were identified from Fedwire funds transactions, Risk Policy to require GSEs and international organizations to as in Furfine (1999). fully fund interest and redemption payments on securities before the funds are sent, and it removed the provision of free Fed funds returns is the total value of the returns of the fed intraday credit to these issuers (Board of Governors of the funds loans. It is equal to the value of fed funds deliveries for Federal Reserve System 2004; McAndrews 2006). the previous business day plus the interest on those loans. These loans were identified from Fedwire funds transactions, MBS P&I day, pre-GSE policy is a binary variable equal to 1 as in Furfine (1999). on the 15th and 25th of the month, or the first business day thereafter, before the change in GSE credit policy on July 1, Payments $10 mn. is the fraction of daily value from 2006. On these days, Fannie Mae and Freddie Mac make payments greater than or equal to $10 million. This excludes interest and redemption payments on mortgage-backed all CHIPS, CLS Bank, DTC, and P&I funding payments. The
108 Changes in the Timing Distribution Appendix A: Data (Continued)
threshold value of $10 million is the value used in a survey CLS Bank opens is a binary variable equal to 1 for all days on of bank intraday liquidity management conducted by the or after September 10, 2002, when CLS Bank International Payments Risk Committee and the Wholesale Customer began settling U.S. dollar transactions. Advisory Group (2007). CLS Bank USD value is the daily sum of payments sent by CLS Number of payments is the daily number of Fedwire funds Bank over Fedwire. It is equivalent to the value of all U.S. dollar payments, including interbank, customer, and fed funds legs settled by CLS Bank. transactions, but excluding all CHIPS, CLS Bank, DTC, and P&I funding payments. Sep. 11-18, 2001, is a binary variable equal to 1 for those dates. This is the period in which the Fedwire payments system was CHIPS settlement at 17:00 — CHIPS settlement time is a disrupted by the terrorist attacks on September 11 binary variable equal to 1 for all days on or after January 18, (McAndrews and Potter 2002). 2000. On that date, the time at which end-of-day CHIPS payouts occurred moved from approximately 16:45 to 17:10. NYSE closures and NYSE early closures — Source:
FRBNY Economic Policy Review / September 2008 109 Appendix B: Settlement Institutions
CHIPS (Bank for International Settlements 1993). CLS Bank began operation in September 2002; as of December 2006, it had CHIPS is a private, large-value U.S. dollar payments system 57 members and settled an average of 290,000 transactions 12 owned and operated by the Clearing House Payments valued at $3.3 trillion per day. Company (Federal Reserve Bank of New York 2002; Bank for CLS Bank uses a payment-versus-payment method in which International Settlements 2003b, 2005). As of April 2007, funds to settle trades are exchanged simultaneously in different CHIPS had 45 members and settled 329,000 transactions currencies. In order to accomplish simultaneous transfers, CLS valued at $1.7 trillion per day.11 From its opening in 1970 until Bank is open during the five-hour settlement window—01:00 2001, CHIPS operated as an end-of-day multilateral net debit to 06:00 EST—when real-time gross settlement systems in settlement system: After CHIPS closed at 04:30 (05:00 after Europe, the Americas, and Asia are open. January 18, 2000), participants with negative net positions would send payments to CHIPS over Fedwire to cover their positions; CHIPS would then send payments to those participants with net positive positions. DTC On January 22, 2001, CHIPS adopted intraday payment finality with a continuous offsetting algorithm to optimize DTC is a securities settlement system that settles the liquidity. All CHIPS participants must fund their accounts with majority of U.S. corporate securities and commercial paper a Fedwire transfer to CHIPS between the opening of Fedwire and transactions. It is a wholly owned subsidiary of Depository 09:00 before they can send or receive payments. These balances, Trust & Clearing Corporation (Bank for International totaling about $3 billion, are used to settle payments during Settlements 2003a, 2005). DTC has 407 participants and CHIPS operating hours. At the close of CHIPS at 17:00, any 86 settling banks. On average, it settles 800,000 transactions unsettled payments are multilaterally netted. These net positions valued at $896 billion per day (Payments Risk Committee are settled over Fedwire via transfers to and from CHIPS. and Wholesale Customer Advisory Group 2007). DTC participants fund their accounts through Fedwire transfers (via a settlement bank for many) to the DTC Federal Reserve account. Money market instruments represent CLS Bank 62 percent of DTC value. The ability of paying agents to accept maturing securities is limited by the agents’ net debit cap. CLS Bank is a payment-versus-payment settlement system To remove the debit cap constraint, agents will make progress that settles foreign exchange transactions in fifteen currencies payments to their accounts via Fedwire transfers to DTC. The (CLS Bank 2007; Miller and Northcott 2002; Bank for majority of this activity occurs between 12:00 and 14:00. At International Settlements 2003a, 2005). CLS Bank is operated 16:00, the DTC settlement process begins. Banks with net by CLS Bank International, a bank-owned Edge Act debits send the net amount to DTC over the net settlement corporation incorporated in the United States. CLS Bank was system at 16:35. At 16:40, DTC sends Fedwire funds transfers founded in response to concerns raised by the G-10 central to participants with net credits (Payments Risk Committee and banks about settlement risk in foreign exchange transactions Wholesale Customer Advisory Group 2007).
12 Source: CLS Bank International (
110 Changes in the Timing Distribution References
Angelini, P. 1998. “An Analysis of Competitive Externalities in Gross ———. 2003. “Expansion of the Operating Hours for the On-Line Settlement Systems.” Journal of Banking and Finance 22, no. 1 Fedwire Funds Service.” Federal Register 68, no. 101: 28826-8. (January): 1-18. ———. 2004. “Policy Statement on Payments System Risk.” Federal ———. 2000. “Are Banks Risk Averse? Intraday Timing of Operations Register 69, no. 187: 57917-31. Docket no. OP-1182. Available at in the Interbank Market.” Journal of Money, Credit, and
Bank for International Settlements. 1993. “Central Bank Payment and ———. 2006. “Consultation Paper on Intraday Liquidity Settlement Services with Respect to Cross-Border and Multi- Management and Payments System Risk Policy.” Federal Currency Transactions.” Committee on Payment and Settlement Register 71, no. 119: 35679-87. Docket no. OP-1257. Systems, CPSS Publication no. 7, September. Basel, Switzerland. CLS Bank. 2007. “CLS Overview.” ———. 2003a. “A Glossary of Terms Used in Payment and Settlement Systems.” Committee on Payments and Settlement Systems. Federal Reserve Bank of New York. 2002. “Fedpoint: CHIPS.” April. March. Basel, Switzerland. Available at
FRBNY Economic Policy Review / September 2008 111 References (Continued)
Koponen, R., and K. Soramäki. 2005. “Intraday Liquidity Needs in a McAndrews, J., and S. Rajan. 2000. “The Timing and Funding Modern Interbank Payment System—A Simulation Approach.” of Fedwire Funds Transfers.” Federal Reserve Bank of New York In H. Leinonen, ed., Liquidity, Risks and Speed in Payment Economic Policy Review 6, no. 2 (July): 17-32. and Settlement Systems—A Simulation Approach. No. E: 31 in Bank of Finland Studies, chapter 3, 73-116. Miller, P., and C. A. Northcott. 2002. “CLS Bank: Managing Foreign Exchange Settlement Risk.” Bank of Canada Review, autumn: Leinonen, H., and K. Soramäki. 2005. “Optimizing Liquidity Usage 13-25. and Settlement Speed in Payment Systems.” In H. Leinonen, ed., Liquidity, Risks and Speed in Payment and Settlement Mills, D. C., and T. D. Nesmith. 2008. “Risk and Concentration in Systems—A Simulation Approach. No. E: 31 in Bank of Finland Payment and Securities Settlement Systems.” Journal of Studies, chapter 4, 117-50. Monetary Economics 55, no. 3, (April): 542-53.
McAndrews, J. J. 2006. “Alternative Arrangements for the Distribution Newey, W. K., and K. D. West. 1987. A Simple, Positive Semi-Definite, of Intraday Liquidity.” Federal Reserve Bank of New York Heteroskedasticity and Autocorrelation Consistent Covariance Current Issues in Economics and Finance 12, no. 3 (April). Matrix.” Econometrica 55, no. 3 (May): 703-8.
McAndrews, J. J., and S. Potter. 2002. “Liquidity Effects of the Events Payments Risk Committee and Wholesale Customer Advisory Group. 2007. of September 11, 2001.” Federal Reserve Bank of New York “Report of the Joint Task Force on the PSR Policy Consultation Economic Policy Review 8, no. 2 (November): 59-79. Paper.” February.
112 Changes in the Timing Distribution Christopher Becher, Marco Galbiati, and Merxe Tudela
The Timing and Funding of CHAPS Sterling Payments
• Participants in CHAPS Sterling often use 1.Introduction incoming funds to make payments, a process known as liquidity recycling. he use of real-time gross settlement (RTGS) systems for Tthe settlement of large-value payments offers considerable • Liquidity recycling can be problematic if advantages, the principal one being the elimination of the credit participants delay their outgoing payments in risk that can arise between participants in deferred net settle- anticipation of incoming funds. ment systems. However, in comparison with deferred net settlement systems, RTGS systems require relatively large • An analysis of CHAPS payment activity shows amounts of liquidity to support payment activity. This liquidity that the level of liquidity recycling, though high, can be sourced from the settlement agent (usually a central is stable throughout the day—a condition bank in the case of large-value payments systems)—in the form attributable to three features of the system. of intraday overdrafts—or from incoming payments from other participants. Obtaining intraday liquidity from a central bank is typically • First, the settlement of time-critical payments costly. In order to minimise this cost and to take advantage in CHAPS supplies liquidity early in the day— of incoming payments as a funding source, participants may liquidity that can be recycled to fund less choose to delay outgoing payments. However, payment delay urgent payments. may itself prove costly. Participants face a trade-off, therefore, between the cost of borrowing from the central bank and the • Second, CHAPS throughput guidelines expected cost of delaying payments. McAndrews and Rajan provide a centralised coordination mechanism (2000) explore this trade-off in a study of payment behaviour that essentially limits any tendency toward in Fedwire. They describe how the use of incoming funds payment delay. to offset outgoing payments allows participants to avoid incurring costly overdrafts from the central bank and hence • Third, the relatively small direct membership reduces the liquidity cost of making payments. Such offsetting of CHAPS facilitates coordination, enabling can be achieved to a greater extent during activity peaks, so members to maintain a constant flux of banks are induced to coordinate their payments around, payments during the day. and thereby to reinforce, these peaks.
Christopher Becher, a policy officer at the European Commission, The authors thank Mark Manning, James McAndrews, participants at was an analyst at the Bank of England at the time this article was written; workshops at the Bank of England, and an anonymous referee for invaluable Marco Galbiati is an economist and Merxe Tudela a senior economist input. The views expressed are those of the authors and do not necessarily at the Bank of England. reflect the position of the Bank of England, the European Commission, the Federal Reserve Bank of New York, or the Federal Reserve System. Correspondence:
FRBNY Economic Policy Review / September 2008 113 This article investigates the factors influencing the timing The analysis is repeated under priced and collateralised and funding of payments in the CHAPS Sterling system, intraday liquidity regimes, as employed in Fedwire and drawing where appropriate on comparisons with payment CHAPS, respectively.1 activity in Fedwire. In the next section, we discuss theoretical Under a collateralised regime, Bech and Garratt find that approaches to the study of payment behaviour and their both early and delayed payments are possible equilibria, application to CHAPS Sterling. The empirical analysis of the depending on the relative costs of liquidity and delay. The timing and funding of CHAPS Sterling payments follows in efficient equilibrium is for both banks to pay early. However, Sections 3 and 4, respectively. Section 5 concludes. for certain levels of delay and liquidity costs, the participants are found to be in a prisoner’s dilemma, in which the dominant
2. Theoretical Studies This article investigates the factors of Payment Behaviour influencing the timing and funding of
Several theoretical studies have addressed the incentives facing payments in the CHAPS Sterling system, participants in RTGS systems. Many focus on the afore- drawing where appropriate on comparisons mentioned trade-off between the cost of liquidity and the with payment activity in Fedwire. expected cost of delaying payments.
strategy for both is to delay payments until the afternoon, even 2.1 Definition of Terms though both would benefit if payments were made in the morning. This incentive to delay arises because it is possible to The measurement of the cost of liquidity varies according to avoid the cost of posting collateral in the morning and instead the regime employed by the settlement agent (in the cases to incur the (cheaper) delay cost. In these cases, there would be described in this article, that agent is the central bank). When a welfare improvement if the participants could be induced credit is supplied unsecured, the cost typically takes the form of to coordinate and to pay earlier. an explicit overdraft fee. Credit may also be provided against Under a regime of priced credit, Bech and Garratt again find eligible collateral, in which case the cost to the participant is the that multiple equilibria are possible. However, in this case, opportunity cost of posting eligible securities with the central participants stand to benefit from synchronising payments bank and hence forgoing alternative uses for those assets. with each other, since no cost is incurred by either participant The cost of delay may take several forms. Financial penalties if payments are “offset” within the time period over which may be incurred for failure to make time-critical payments overdraft fees are calculated. The equilibrium outcome will by specified deadlines, such as for settlement payments in thus depend not only on the relative costs of liquidity and delay ancillary systems or repayments of interbank loans. In but also on the likelihood that the other bank will receive a addition, failure to make customer payments on time, or payment request. In the specific case where the expected cost of indeed at all, on the intended settlement date may result in delay is lower than the credit fee, and payment flows are skewed reputational costs and a loss of future business. Also, as we toward the afternoon, Bech and Garratt find that the efficient discuss later, the reputation of a participant within a payments equilibrium involves delay until the afternoon. system may suffer if it is perceived to be delaying payments In a similar study, Kobayakawa (1997) models the choice of in order to “free-ride” on liquidity provided by others. whether to delay payments in RTGS systems under varying intraday credit arrangements. Again, the relative costs of liquidity and delay drive equilibrium selection. Under a system of priced credit, Kobayakawa (like Bech and Garratt) finds that 2.2 Theoretical Approaches delayed settlement is an equilibrium, since each participant seeks to avoid incurring an overdraft by delaying payments and Bech and Garratt (2003) model the trade-off using a game- thereby “free-riding” on the liquidity provided by the other theoretical approach, analysing the behaviour of two banks, participant. Under a collateralised regime, Kobayakawa finds both of which receive random payment requests from a unique equilibrium in which both participants pay early. customers at the beginning of a morning and an afternoon period. Both banks face a fixed cost of delaying payments 1 Bech and Garratt also examine the case of free intraday credit; however, those and of posting collateral for a morning or afternoon period. results are not discussed here.
114 The Timing and Funding of CHAPS Sterling Payments However, in this case, the result is obtained by assuming that of delay and of overdrafts), on the time-criticality of payments, the opportunity cost of collateral is a sunk cost, and so liquidity and on the probability and size of liquidity shocks, the resulting is in effect free when the game is played. This is unsatisfactory, equilibria feature different degrees of delay. More specifically, since it takes no account of the participants’ incentives to some or all banks, depending on the shocks received, decide to reduce the cost of liquidity by economising on the value of delay their payments. Martin and McAndrews also explore the collateral posted. Consequently, we focus in the following effect of a liquidity-saving mechanism on the banks’ incentives. discussion on the Bech and Garratt model. This is shown to mitigate the strategic complementarity of Mills and Nesmith (2008) adapt the Bech and Garratt model banks’ strategies by allowing banks to release payments to look at the effect of settlement risk on timing decisions in conditional on the receipt of payments. The paper shows that, payments and securities settlement systems, concentrating on in this case, the extent of delay in equilibrium also depends on the differential impact of overdraft costs on the two types of the pattern of payments (whether payments can be offset in systems. A main contribution of this paper is to describe a pairs or multilaterally), which has implications for the system’s rationale for delays overlooked in the literature: namely, that efficiency. banks may withhold payments until they receive information The overarching conclusion of these works is that institu- on the others’ ability to send funds, in order to obtain a better tional features and, in particular, intraday credit regimes have forecast of the costs of funding their own payments. More a powerful effect on banks’ incentives; as a consequence, they precisely, in the model, banks choose between paying “early” largely determine a payments system’s performance. We draw and paying “late.” There are no delay costs, so, in the absence on—and, where necessary, modify—these theoretical of settlement risk, “early” is a weakly dominated strategy,2 predictions to study payment behaviour in Fedwire and which may still appear in equilibrium (although only in in CHAPS Sterling. risk-dominated ones). Introduction of settlement risk definitely tilts the balance against the “early” strategy, because 2.3 Implications for Payment Behaviour in Fedwire The overarching conclusion of these [theoretical] works is that institutional The Federal Reserve System supplies intraday liquidity to features and, in particular, intraday credit Fedwire members in the form of uncollateralised daylight overdrafts. Subject to net debit caps, participants can incur regimes have a powerful effect on banks’ overdrafts at any time, which incur a charge calculated as incentives; as a consequence, they largely the average per-minute overdraft during the day, multiplied by an effective daily rate, less a deductible.3 determine a payments system’s As Bech and Garratt (2003) describe, a corollary of this performance. charging structure is that participants can avoid overdraft charges by synchronising payments. As long as incoming payments of at least equivalent value to outgoing payments in this case “early” becomes a strictly dominated strategy are received within a minute, no overdraft will be required (“late” outperforms it against any action by the opponent, and hence no charge incurred. This sets the scene for a as settlement risk imposes some overnight overdraft in pure coordination game, in which participants attempt to probability terms). Thus, settlement risk introduces further synchronise payments in order to minimise the average reasons to delay, eliminating the “early payment” equilibria. overdraft position over the course of the day. The theory would Building on previous work, Martin and McAndrews (2008) also predict that if the overdraft fee is deemed to be high construct a model with a continuum of banks, each making relative to the expected cost of delay, the efficient equilibrium a unit payment to one other bank and each having to decide will involve the delay of payments until later in the day. whether to pay “early” or “late.” Banks are assumed to face This theoretical finding is consistent with the empirical random delay and liquidity costs, determined in turn by bank- results obtained by McAndrews and Rajan (2000) on the specific shocks that drain (or increase) the available liquidity. timing and funding of payments in Fedwire. Faced with costly The paper shows that, depending on the cost parameters (costs intraday liquidity, participants appear to delay payments until an end-of-day activity peak, during which the probability of 2 “Early” performs no better than “late,” and it performs just as well as “late” if the other also pays early, as offsetting payments incur no charge. 3 The effective rate is currently equivalent to an annual rate of 36 basis points.
FRBNY Economic Policy Review / September 2008 115 receiving funds from other participants is greater. It is argued therefore face an opportunity cost. As described in Box 1, the that this synchronised delay reinforces the activity peak. The upper bound to this cost has been estimated to be of the order “focal points” for this coordination appear to be provided by of 7 basis points per annum, although for domestic banks ancillary system settlement deadlines (in particular, in CHIPS subject to the Stock Liquidity Regime the opportunity cost may and DTC). As McAndrews and Rajan note, though, the be significantly lower and may even approach zero. outcome of this apparent coordination may not be socially The Bech and Garratt (2003) model predicts that this regime efficient, since all participants might stand to benefit from will result in multiple equilibria, with the selection of an reduced liquidity costs if coordination could be improved equilibrium dependent on the relative magnitudes of the cost so as to take full account of liquidity externalities. of delayed payment and the opportunity cost of posting Armantier, Arnold, and McAndrews (2008) extend this collateral. The low opportunity cost of posting collateral for analysis in their study of recent changes in the timing of Fedwire funds transfers. Among other things, they explore alternative hypotheses for the timing of late-afternoon In contrast with Fedwire, where the total payment peaks, with particular reference to the change in the cost of liquidity is driven by the average timing of late-afternoon Fedwire transfers following a move to a later CHIPS settlement time. The tendency for Fedwire overdraft incurred, the cost of liquidity in transfers to be made after ancillary system positions are CHAPS Sterling is driven by the maximum settled may reflect the “focal point” hypothesis described overdraft position incurred during the day, above. However, it may also be that the liquidity released by ancillary system settlement may trigger a “cascade” of since the value of collateral posted must payments, to the extent that participants are liquidity- be at least equal to this position. constrained.4 Along similar lines, the settlement may also release credit lines, thereby permitting more payments to be made. Additionally, the authors suggest that uncertainty many CHAPS Sterling members may thus be expected to surrounding the size of ancillary system payouts may lead to favour an early rather than a delayed payment equilibrium. payments being deferred until after the settlement deadline— That said, it is difficult to quantify the cost of delay associated that is, once uncertainty has been resolved. The data do with all but a small number of time-critical payments. not allow for a clear distinction to be made between the Anecdotal evidence suggests that costs of delay are low for competing hypotheses; however, there is sufficient evidence the majority of payments. to suggest that the coordination described in the earlier paper Certain qualifications are required in applying this model to is only part of the story. CHAPS Sterling. In particular, in the Bech and Garratt model, the benefit from delaying payments in a collateralised regime derives from the assumption that it is less costly to post collateral for the afternoon than for the whole day (and hence 2.4 Implications for Payment Behaviour in CHAPS Sterling that there is an incentive to avoid posting collateral in the morning). This in turn rests on the assumption that it is possible to invest surplus liquidity for a fraction of the day—or, The Bank of England provides intraday liquidity to members of CHAPS Sterling in the form of interest-free overdrafts secured in other words, that there exists an intraday market for against eligible collateral. The maximum value of liquidity liquidity. It is not obvious that this incentive applies in CHAPS granted is equal to the value of collateral securities posted, less Sterling since, in the absence of an intraday market, it is a “haircut” to take account of movements in the value of the probably no cheaper to post collateral for a morning or collateral securities. In contrast with Fedwire, where the total afternoon than for a full day. Once collateral is committed to 5 cost of liquidity is driven by the average overdraft incurred, the the payments system, the cost for the full day is incurred. cost of liquidity in CHAPS Sterling is driven by the maximum It is possible to modify the Bech and Garratt model to overdraft position incurred during the day, since the value incorporate an incentive to delay that does not rely on the of collateral posted must be at least equal to this position. existence of an intraday market for liquidity. By delaying The cost of posting collateral derives from the fact that the 5 It is nonetheless possible for banks to withdraw liquidity intraday. As we securities posted (or the funds used to obtain the required argue here, while it may not be possible to lend in an intraday interbank securities) could be used for alternative purposes; participants market, the collateral could in principle be committed to another payments system. In this case, the ability to commit collateral for only part of a day could 4 See also Beyeler et al. (2006). be considered valuable.
116 The Timing and Funding of CHAPS Sterling Payments Box 1 payment and taking advantage of incoming funds, a The Cost of Liquidity in CHAPS Sterling participant may be able to reduce the maximum overdraft position and hence reduce the aggregate collateral requirement James and Willison (2004) estimate the opportunity cost of posting for the day (or avoid posting collateral altogether). It can be eligible collateral as the difference between the unsecured shown that a similar prisoner’s dilemma outcome emerges interbank rate and the secured-lending repo rate. By posting from this model, with both participants defecting despite collateral, the bank forgoes the opportunity to repo the securities the mutual benefit from paying early, unless they can somehow and to lend the funds obtained at a higher rate in the interbank be induced to coordinate earlier in the day.6 market. However, the finding that some participants may seek to As James and Willison acknowledge, this is only part of the reduce their aggregate collateral requirements by delaying story. U.K. banks are subject to the Stock Liquidity Regime (SLR), payments must be seen in the context of the empirical under which they are required to hold liquid assets sufficient to observations that participants in CHAPS typically post cover net sight deposit and five-day wholesale cash outflows. collateral at the beginning of the day (that is, they do not These assets cannot be repurchased overnight to generate cash generally wait to determine whether collateral posting is in the interbank market, but they can be posted with the Bank required) and many post collateral to a value well in excess of of England to generate liquidity in CHAPS Sterling (since the liquidity usage (for the system as a whole, maximum liquidity SLR requirements are measured only at the end of the day). For used is only around one-third of the maximum collateral those banks subject to the SLR, the opportunity cost of posting posted).7 Two factors appear to be particularly influential collateral—and hence the cost of liquidity—may be even lower in explaining this behaviour.8 than the 7 basis point estimate. First, as discussed in Box 1, the low opportunity cost of Three foreign settlement banks in CHAPS Sterling— posting collateral (or of maintaining positive reserve account accounting for around 14 percent of transactions by value or balances) means that, for many banks there appears to be 11 percent of transactions by volume—are not subject to the SLR. little incentive to delay posting collateral until later in the day, These banks are instead subject to the Maturity Mismatch Regime, since the potential savings to be made from reducing the which does not require banks to hold eligible liquid assets if aggregate value of collateral posted are small. committed outflows equal expected inflows. For these banks, Second, the distribution of “time critical” payments, for the opportunity cost of posting collateral may be higher. However, which the expected cost of delay is high, appears to be skewed the use by these banks of cross-border collateral arrangements toward the morning in CHAPS Sterling. For example, pay-ins (including the ability to back sterling payments with euro cash to CLS Bank must be made by 11:00 a.m. in order to avoid collateral) implies that estimation of the cost of liquidity for significant financial penalties, and market convention dictates foreign banks would require analysis of the cost of generating that overnight interbank loans should be repaid the following liquidity in other jurisdictions. morning. Even for those banks for which collateral posting is Following reform of the Bank of England’s money market relatively costly, the expected cost of delay for time-critical operations in May 2006, participants are now able to hold payments may be so high as to warrant posting sufficient remunerated reserve account balances with the Bank of England. collateral at the beginning of the day to ensure that liquidity is These balances can be used to fund payments, so members can available to make these payments without the need for recourse choose to provide liquidity in this way rather than by posting to incoming funds. The existence of throughput guidelines eligible securities. The relative opportunity cost of holding reserve may serve to reinforce the incentive to post liquidity “up front” account balances may vary from member to member, although (as we discuss in more detail in Section 2.5). for some foreign banks, particularly those that do not routinely hold sterling collateral, reserve balances may be a relatively 6 We are indebted to Peter Gibbard for these insights. attractive source of liquidity. 7 It should be noted that CHAPS Sterling payments are not the only claim on the available liquidity. Participants in the United Kingdom’s securities settlement system, CREST, are able to transfer liquidity from CHAPS Sterling settlement accounts to separate accounts designated for the settlement of the cash legs of securities transactions. Liquidity is also available for the settlement The value of collateral posted to support these time-critical of positions in other ancillary systems, such as CLS Bank, retail payments payments depends on the expectation of the size of time- systems (BACS and C&CC), and LCH.Clearnet. Unlike CLS Bank pay-ins, the latter transfers do not take place in CHAPS Sterling and therefore are not critical payment flows. Consider a single period in which delay recorded in the payment data in this article. costs are zero up to a certain time (say, the deadline for a 8 In addition to the factors described, there may be frictions associated with time-critical payment) and very high thereafter. If all payment obtaining eligible securities during the day that tend to encourage early instructions are known at the beginning of the period in which posting. Under normal circumstances, intraday repos with the Bank of England are unwound at the end of the day and the securities are held in custody by time-critical payments must be made—and therefore all banks the Bank of England overnight, to be reposted the following morning. are aware of whether they will be net payers or net receivers at
FRBNY Economic Policy Review / September 2008 117 the end of the period—it can be shown that each net payer Banks’ incentives are therefore aligned and consistent with must raise liquidity at least equal to the value of its net the efficient use of liquidity. payments and that this amount will be necessary and sufficient However, this result requires that all participants know at the to settle all payments in the system (Box 2). Net receivers will beginning of the period whether they will be net payers or net be able to meet their payment obligations using incoming receivers at the end. In reality, participants will not typically funds and hence will not need to raise additional liquidity. possess full information about payment flows at the time when collateral posting decisions are made, and hence they will face uncertainty about liquidity requirements. Faced with high
Box 2 threshold delay costs, participants will wish to insure themselves Liquidity Requirements for Time-Critical Payments against the risk of being net payers and—if the cost of failing to make a time-critical payment is sufficiently high—they may For simplicity, assume that banks receive all payment instructions choose to post liquidity at the beginning of the period to a value exogenously from their customers. These instructions are denoted at least equal to the maximum anticipated gross value of the ij by x t —that is, at time t, bank i is requested by its customer(s) to time-critical payments. For payments with low delay costs, pay the amount x to bank j. A payment from bank i to bank j at by contrast, participants may be willing to rely on incoming ij time t is denoted pt . Banks choose whether to settle payment funds rather than post additional collateral. We illustrate this ij instructions immediately or to queue them internally. So, pt scenario using a simple stylised framework (see exhibit). ij and x t need not be the same. Here, the choice of the value of collateral posted at the Consider the case in which delay costs are zero up to a certain beginning of the day is determined by the intersection of time T and subsequently so high that all payments must be settled the expected cost of delay (which varies across payments; in the
within T. It follows that: exhibit, X1 percent of payments incurs a delay cost of at least X) ij ij ∀ , ∑ x t = ∑ pt ij. and the cost of collateral (which, as described above, is fixed t =0…T t =0…T once collateral is posted). The value of collateral posted must The payment balance of bank i against bank j at time t is be sufficient to ensure that time-critical payments—for which defined as: the cost of delay is greater than the cost of liquidity—can ij ji ij bt = ∑ ps –∑ ps . be made without the need for recourse to incoming funds. s=0…t s=0…t By contrast, for those payments for which the cost of delay Bank i is a net payer for the period if its total payment balance is lower than the opportunity cost of posting collateral (the at T is negative—that is, if b i = ∑ bij<0 . Since customer T ji≠ T proportion of payments 100 percent minus X1 for participant orders are exogenous, banks cannot affect whether they will be net X), participants may be willing to rely on the recycling of payers or net receivers. We assume, however, that banks know with incoming funds instead of posting additional liquidity. Such certainty at the beginning of the period which type they will be. payments—particularly those of high value—will typically be We define I as the set of net payers. Each net payer i has delayed until after time-critical payments are settled, especially i ∑ ij to raise liquidity to a value at least equal to L = – ji≠ bT in when there is uncertainty over the liquidity demands of time- order to execute its payment instructions. Hence, ∑ Li is the iI∈ minimum liquidity required to settle all payment instructions. This amount will also be sufficient to settle all payments by Delay and Liquidity Costs in CHAPS Sterling System
time T if, first, every net payer raises Li at time zero and pays it out immediately and, second, at any t, every bank i uses all of its Percentage of payments liquidity to make payments up to that value (or less, up to the exhaustion of queued orders).a Because delay costs are zero up to 100 Low collateral cost T, this pattern is optimal for all banks. We can therefore conclude that, when delay costs are zero up to a time-critical threshold and very high thereafter, the banks’ interests are aligned and High collateral cost compatible with the efficient use of liquidity. All payments
are settled using only the minimum liquidity L. Y1 Percentage of payments imposing (at least) a We are abstracting here from 1) the indivisibility of payments—that is, a given delay cost additional liquidity may be required if payments cannot be split and settled X1 in tranches, and 2) the possibility that no bank is a net payer—meaning that all payments net out exactly. In this case, a bargaining process would be required to define who is to provide liquidity, given that liquidity is required, yet no one needs to post collateral if somebody else does. Y X Delay cost
118 The Timing and Funding of CHAPS Sterling Payments critical payments.9 In addition, to the extent that there is models. However, we have also seen that for payments for reliance on incoming funds, the liquidity released by the which the cost of delay is relatively low, participants may rely settlement of time-critical payments may result in a “liquidity to a greater extent on incoming funds as a funding source in cascade,” as queued payments are released. In this way, the order to avoid squeezing the precautionary buffer of spare early settlement of time-critical payments may serve to catalyse liquidity—or indeed to avoid posting additional collateral. liquidity recycling later in the day. The efficiency with which incoming funds are recycled The proportion of payments to which this applies will vary will depend on the extent to which participants collectively according to the cost of liquidity. As the exhibit illustrates, maintain the flow of liquidity around the system, perhaps via banks for which the opportunity cost of posting collateral is proactive payment coordination. McAndrews and Rajan relatively low (for example, cost of collateral Y) will post a larger stock of collateral and hence may tend to fund a greater The efficiency with which incoming funds volume of payments (Y1) from posted liquidity than from incoming funds. However, banks for which liquidity is are recycled will depend on the extent to relatively costly may post less collateral and remain more which participants collectively maintain reliant on the recycling of incoming payments to fund outgoing payments. the flow of liquidity around the system, Of course, as in the single-period example above, we must perhaps via proactive payment also take account of the limited information available to participants when decisions are made. While a proportion of coordination. payment instructions may be known at the start of the day (which we would expect to be relatively large when payment (2000) describe how such coordination is achieved in Fedwire activity is driven by proprietary rather than customer through the delay of payments until an end-of-day activity business), there may remain considerable uncertainty about peak, but suggest that the observed level of coordination may the size and distribution of incoming and outgoing payments, be inefficiently low—and hence liquidity costs inefficiently including payments with high delay costs. Faced with such high—due to collective-action problems, exemplifying the uncertainty about aggregate liquidity demands, participants uncooperative outcome in the prisoner’s dilemma game would be expected to hold a buffer of liquidity in excess of the described by Bech and Garratt (2003). In principle, CHAPS quantity predicted in this framework in order to withstand Sterling members could suffer from a similarly uncooperative unforeseen liquidity demands. The tendency to maintain such outcome in which some members defect and withhold liquidity cushions—which are indeed observed in practice— payments, thereby curtailing the ability of others to take will also contribute to the entire system’s resilience to liquidity advantage of incoming funds. shocks, such as the operational failure of one or more banks to In practice, the early settlement of time-critical payments make payments. will contribute to the recycling of liquidity by ensuring that payments begin to flow early in the day. In addition, certain features of CHAPS Sterling may be particularly conducive to achieving a cooperative outcome and hence to ensuring that 2.5 Liquidity Recycling in CHAPS Sterling liquidity is recycled efficiently. For example, the CHAPS Clearing Company imposes a set of throughput guidelines, We have discussed how the apparently low opportunity cost of whereby participants are expected to make 50 percent of collateral for many participants and the high expected delay payments by value by 12:00 p.m. and 75 percent by 2:30 p.m., costs associated with a subset of payments will tend to favour as an average over a calendar month, with the explicit intent posting collateral at the beginning of the day. This may serve to of improving the efficiency of liquidity usage in the system. reduce the incentive for payment delay and hence avoid the While enforcement of the guidelines relies on peer pressure prisoner’s dilemma outcomes described in the theoretical rather than legal compulsion (Box 3), the guidelines are largely 9 Empirically, banks tend to settle a large volume of low-value payments early observed in practice. As Buckle and Campbell (2003) in the day. This might appear to contradict the predictions of this model, since demonstrate, the existence of such guidelines acts to the inherent delay cost associated with any one of these payments is likely to be countervail any tendency toward payment delay and hence low. However, expected delay costs for these payments collectively may be high, since processing of large volumes later in the day may prove difficult and costly serves to promote liquidity recycling earlier in the settlement in the event of an operational incident. day, thereby enhancing the efficiency of the payments system.
FRBNY Economic Policy Review / September 2008 119 Box 3 Enforcement of CHAPS Throughput Guidelines punishment strategy, such as delaying their own payments to that member. This cost associated with being perceived If a CHAPS Sterling member breaches the throughput guidelines as “free-riding” on liquidity provided by peers in a repeated in three consecutive months, that member is required to provide game may thus induce a cooperative outcome. reasons to the CHAPS Clearing Company and to outline the One specific mechanism possibly enforcing such discipline steps taken to ensure that deadlines are met going forward. The is the use of bilateral net sender limits. This is the simple participant will be given the opportunity to provide evidence that, liquidity management rule whereby bank A ceases to make over the period in question, failure to meet the guidelines resulted payments to bank B if the net flux of payments from A to B from a lack of payment instructions rather than a shortage of reaches a certain (positive) limit; in other words, B is available liquidity. “punished” if it is seen “not to reciprocate.” Anecdotal If the member breaches the guidelines in six consecutive evidence suggests that this mechanism is indeed applied months, or in three consecutive months on two occasions, and has by some CHAPS members. The appendix formalizes the been unable to provide evidence as set out above, it will be obliged argument, but the logic behind bilateral net sender limits is to attend a “Star Chamber” hearing. At the Star Chamber, the that they create “interperiod spillovers,” increasing the cost member’s CHAPS board director will be required to explain the of delaying payments by depriving the recalcitrant bank steps being taken to resolve the issues and to return performance of liquidity in subsequent periods. As a result, banks are to acceptable service levels and guidelines. encouraged to make payments promptly, to the benefit of the 12 There is no defined penalty for the breach: As a rule, peer system. As shown in the exhibit, the effect of such limits is to pressure is felt to be sufficient. However, the CHAPS Rules give shift the delay curve to the right. In this framework, the result the company manager the power to suspend or exclude a member would be to increase the value of collateral posted at the “in material breach” of the provisions of the procedural rules, beginning of the day. The effect on liquidity usage would or where, in the opinion of the CHAPS Clearing Company, depend on the effect on liquidity recycling over the course circumstances have arisen that could be “prejudicial” to the system of the day. or represent a threat to its “security, integrity, or reputation.” But even though centralised throughput guidelines and decentralised mechanisms may serve as coordination devices, the use of bilateral net sender limits (in the form described above) would not enforce coordination on a particular time In addition, the concentrated structure of CHAPS Sterling of the payment day and hence would not necessarily overcome appears more conducive to coordinated behaviour than does a tendency to delay payments until the end of the day. Acting the structure of Fedwire, which has a broader membership. in tandem, however, throughput guidelines and bilateral CHAPS has fifteen direct members (including the Bank of coordination mechanisms can be expected to both enhance the England), and the majority of payments are made by a core efficiency of liquidity recycling and to smooth the intraday of four participants. The Fedwire network is more extensive, distribution of payments. We seek evidence of these effects as around 9,500 participants access the clearings directly.10 in the empirical analysis that follows. This results in very different network topologies, which in turn has implications for the flow of liquidity around the systems.11 In particular, the concentration of payment flows among a small group of banks in CHAPS Sterling leads naturally to 3.The Timing of CHAPS a higher level of recycling throughout the day than would occur Sterling Payments in a more dispersed system, since each unit of liquidity paid out is more likely to be returned quickly if payments are flowing We now turn to an empirical analysis of the timing and funding between fewer banks. Furthermore, within a “small club” of payments in CHAPS Sterling. Based on the discussion above, of participants, the behaviour of each participant is highly we would expect the low opportunity cost of posting collateral visible to others. If one participant defects and fails to provide and the high expected delay costs associated with a subset liquidity to the system, other participants may adopt a of payments to limit the degree to which members delay payments in order to take advantage of incoming funds. We 10 Only a small proportion of these banks use Fedwire heavily, however. 11 The network of payments between settlement banks in CHAPS, however, 12 Bilateral limits also have the important function of reducing the impact of is underlain by a more extensive network of payments between the originators “liquidity sinks,” created when a bank is able to receive payments but is unable of payments and the end recipients. The characteristics of this network are to release funds—for example, as a consequence of a technical outage. By similar to those of Fedwire. See Soramäki et al. (2006) and Becher, Millard, restricting flows to the “sink” bank, bilateral limits reduce the amount of and Soramäki (2007). liquidity that is syphoned out of the system.
120 The Timing and Funding of CHAPS Sterling Payments Chart 1 Chart 2 Value and Volume of Payments of All Banks Value and Volume of Payments of Foreign Banks in CHAPS Sterling System in CHAPS Sterling System
Billions of British pounds Number of transactions Millions of British pounds Number of transactions 9 4,000 1,400 800 Value Volume Volume 8 Scale Scale 3,500 1,200 Scale 700 7 Value 3,000 600 1,000 Scale 6 2,500 500 5 800 2,000 400 4 600 1,500 300 3 400 200 2 1,000 200 1 500 100 0 0 0 0 06:00 08:00 10:00 12:00 14:00 16:00 06:00 08:00 10:00 12:00 14:00 16:00 Time Time Sources: CHAPS payment database; Bank of England calculations. Sources: CHAPS payment database; Bank of England calculations. Note: We calculate the figures as daily average values and volumes Note: We calculate the figures as daily average values and volumes in ten-minute intervals, using data from October 2006. in ten-minute intervals, using data from October 2006.
would expect to see this reflected in the intraday payment Chart 3 distribution. We also hypothesise that structural features of Value and Volume of Payments of Domestic Banks the CHAPS Sterling system, in particular the imposition of in CHAPS Sterling System throughput guidelines and the “small club” membership, will Billions of British pounds Number of transactions promote the recycling of liquidity and smooth the distribution 9 4,000 Value Volume of payments throughout the day. 8 Scale Scale 3,500 We first consider the intraday pattern of payments in 7 CHAPS Sterling. The system opens for normal service at 3,000 6 6:00 a.m. and closes at 4:20 p.m. CHAPS settlement banks can 2,500 initiate transfers on behalf of themselves and their clients 5 2,000 normally until 4:00 p.m., although settlement members may 4 1,500 make transfers on their own behalf, or on behalf of other credit 3 institutions and certain money market participants, for the 2 1,000 purpose of settling their end-of-day positions after this time. The intraday profiles of payments in CHAPS Sterling are 1 500 shown in Charts 1-3, alongside the profile of Fedwire payments 0 0 13 06:00 08:00 10:00 12:00 14:00 16:00 in Chart 4. The profile of payments by all banks (Chart 1) Time displays three distinct value peaks: the first around 9:30 a.m.; Sources: CHAPS payment database; Bank of England calculations. the second between roughly 11:30 a.m. and 12:00 p.m.; and Note: We calculate the figures as daily average values and volumes a third, sustained peak between 2:00 p.m. and 4:00 p.m. in ten-minute intervals, using data from October 2006. The value profiles are similar for foreign and domestic banks (Charts 2 and 3, respectively: on average, 55 percent of foreign bank payments are made by noon, compared with 58 percent for domestic banks. Both domestic and foreign banks make 13 The analysis in this section is based on data for CHAPS Sterling payment around 25 to 30 percent of payments by value during the flows only; other transfers, such as settlement payments for BACS and C&CC, end-of-day value peak. The profile contrasts with that of are not included. Our results apply to one month only (October 2006); however, our analysis has been repeated for data from June 2005 and payments in Fedwire, which exhibits strong concentration January 2006, with similar results. of payment value at the end of the day (Chart 4).
FRBNY Economic Policy Review / September 2008 121 Chart 4 Chart 5 Value of Payments Made by Time of Day in Fedwire Effect of Payment Events on the Intraday Distribution Daily Average by Month of CHAPS Sterling System Payments
Billions of U.S. dollars Billions of British pounds 250 9 April 2007 8 200 CHAPS 7
6 150 5
4 100 April 2006 3 Loan 50 2 Loan advances BACS repayments 1 CLS Bank C&CC April 2005 0 0 22:00 0:00 2:00 4:00 6:00 8:00 10:00 12:00 14:00 16:00 18:00 06:00 08:00 10:00 12:00 14:00 16:00 Time Time
Source: Federal Reserve Bank of New York. Sources: CHAPS payment database; Bank of England calculations.
Note: We calculate the figures as daily average values and volumes in ten-minute intervals, using data from October 2006.
Setting aside for the moment the effect of strategic behaviour, we note that the observed value peaks correspond well with scheduled payment events, particularly those value peaks are likely to be reinforced by the creation of associated with time-critical payments and throughput overnight loans for the purposes of position-squaring, which deadlines. This result is illustrated in Chart 5. must typically be repaid the following morning. Going The first peak (9:30 a.m.) temporally corresponds both with forward, the effect of overnight markets on payment profiles the timetable for pay-ins to CLS Bank, which can be made and liquidity usage is fertile ground for future research. during a payment window between 7:00 a.m. and 11:00 a.m., The volume profile of CHAPS Sterling payments is relatively and with the settlement of multilateral net positions in the smooth throughout the day for the system as a whole. Volume BACS retail payments system.14 In both cases, the value of the peaks occur shortly after opening and again late in the day. The settlement payments involved is small relative to the total value volume profile is notably different for the set of foreign banks: of payments made at these particular times. However, the Volumes are highly concentrated during the first two hours observed peak may reflect the tendency to delay payments until and fall away sharply thereafter. Approximately 40 percent of after the time-critical payments have been made, when any foreign banks’ payments by volume are made by 8:00 a.m., uncertainty around the value of these settlements has been resolved. The settlement payments may also release liquidity compared with only around 15 percent for domestic banks. for the settlement of subsequent payments. Sharp value peaks This may in part reflect the settlement of payments queued also occur ahead of the throughput deadlines, at noon and between the opening of continental European markets and 2:30 p.m., suggesting that the guidelines do impact significantly the opening of CHAPS Sterling. on the intraday distribution of payments. In fact, there is prima The concentration of payment value in Fedwire appears facie evidence that payments are delayed until the period to be driven by a distinct skew in large-value payments toward immediately before the deadlines, which may reflect strategic the end of the day (Chart 6). This distribution may reflect behaviour.15 institutionally imposed timings for certain types of payments, Finally, the value peaks may also reflect the routine patterns such as for CHIPS and DTC settlement; the creation of of activity in the overnight interbank market: Late-afternoon overnight loans; and settlement payments in financial markets. This is consistent with McAndrews and Rajan’s (2000) 14 The plots shown do not include these payments. observation that this peak existed prior to the imposition of 15 The noon peak also follows the settlement of positions in the C&CC. overdraft fees. However, as discussed above, the peak may This may be influential, although the very low value of settlement payments in this system suggests that it is unlikely to trigger a liquidity cascade or additionally serve as a focal point for, and be reinforced by, to generate material uncertainty for participants. proactive payment coordination.
122 The Timing and Funding of CHAPS Sterling Payments Chart 6 Chart 7 Distribution of Fedwire Payments Distribution of CHAPS Sterling System Payments by Size of Payment by Size of Payment One-Minute Intervals, Daily Average, October 2006 Number of transactions Thousands of transactions 450 15 Payments greater Number of transactions Number of transactions 400 than or equal to 600 Payments 9 $100 million less than 99 pc 8 350 Scale Scale 500 Payments 7 10 greater than 99 pc 300 Scale 400 6 250 Payments less than $100 million 5 200 Scale 300 4 150 5 200 3 100 2 50 100 0 0 1 22:00 0:00 2:00 4:00 6:00 8:00 10:00 12:00 14:00 16:00 18:00 0 0 Time 06:00 08:00 10:00 12:00 14:00 16:00 Time Source: Federal Reserve Bank of New York. Sources: CHAPS payment database; Bank of England calculations. Note: $100 million was the 99th percentile for payment size on March 19, 2007.
payments are much less concentrated at the end of the day Large-value payments in CHAPS Sterling, defined as those in CHAPS Sterling—provides some initial evidence that the that fall within the 99th percentile of the distribution of incentives for payment delay are weaker in CHAPS Sterling payment sizes, account for around 75 percent of daily payment than in Fedwire.17 The profile of CHAPS Sterling payments is value. As Chart 7 illustrates, the volume of payments per clearly influenced by the existence of time-critical payments minute in this category is very small, and the intraday and throughput guidelines. To consider whether these patterns distribution is smoother than in Fedwire (Chart 6). Peaks are also influenced by the strategic behaviour of members, in the incidence of large-value payments occur at 9:30 a.m., we now attempt to disaggregate the sources of funding. In 12:00 p.m. (perhaps related to the first throughput guideline), particular, we assess whether there is evidence that the use and at the end of the day (consistent with the timing of large of incoming funds varies by time of day and, following position-squaring payments in the interbank market). This McAndrews and Rajan (2000), whether peaks in liquidity suggests that the distribution of large-value payments is driven recycling coincide with peaks in payment activity. by the purposes of the payments in question and less by a generalised tendency to delay until the end of the day. The early concentration of time-critical payments in CHAPS Sterling (in conjunction with other ancillary system 4.The Funding of CHAPS settlement payments and the possible need for additional Sterling Payments liquidity transfers to CREST) may also help explain why the average value of payments made during the first two hours of opening is low. Participants may be reluctant to commit liquidity to other large-value payments until these time- 4.1 Methodology critical payments have been made. This applies less forcefully to low-value payments; indeed, it is apparent from Chart 7 To decompose the sources of funding of CHAPS Sterling that low-value payments are released into the system early, payments, we distinguish between two sources of funding: perhaps reflecting their low consumption of liquidity.16 1) payments received from other CHAPS Sterling participants The contrast between the payment profiles in CHAPS within a specified time interval and 2) account balances held at Sterling and in Fedwire—in particular, the observation that the Bank of England, funded both by collateral posting and the maintenance of positive reserve account balances. 16 The tendency to settle a high volume of low-value payments early in the day may also reflect the relative complexity of settling high volumes of low-value 17 Of course, it is not possible to observe delay directly from the intraday payments in the event of an operational disruption later in the day. See also payment profile. This would require knowledge of the timing of payment footnote 9. instructions, as well as of settlement.
FRBNY Economic Policy Review / September 2008 123 Box 4 Measurement of “Offsetting” Payments 4.2 Results in CHAPS Sterling We now decompose gross payments into the constituent We define the value of payments made from member i to funding sources (Chart 8). While the absolute value of t member j within minute t as pij . payments “offset” by incoming payments increases when the The total value of payments made within that minute is value of gross payments increases, the proportion of payments t therefore ∑ pij and the value of net payments for each funded by incoming payments remains comparatively stable ij, throughout the day. On average, during the day, around ()t t member, i, is ∑ pij – pji = N . 23 percent of payments made are funded using incoming j payments; the use of incoming payments peaks at around The value of payments not offset within a minute is equal to the sum of net payments for the set of members for which N is positive, 42 percent shortly before noon. Compared with the results for Fedwire (Chart 9), CHAPS or, equivalently, -1- N . 2 ∑ Sterling does not exhibit a pronounced peak in the share of i incoming payments as a funding source during the end-of-day The value of offsetting payments is then calculated as the value value peak, although the proportion of payments funded by of gross payments made, less the value of payments not offset: incoming funds is above the daily average at this time. It is, ∑ p t – -1-∑ N . ij 2 however, notable that a distinct peak occurs shortly before ij, i the first throughput deadline (at 12:00 p.m.), suggesting that incoming payments are a particularly important funding source at this time. This may ease the liquidity demands of In their study of payment activity in Fedwire, McAndrews meeting the throughput deadline; indeed, this concentration and Rajan (2000) consider incoming payments as a funding may be a product of deliberate payment coordination on the source for outgoing payments only if those incoming payments focal point(s) provided by throughput guidelines. offset outgoing payments within the same one-minute interval. Perhaps the most striking observation from Charts 8 and 9, This definition follows naturally from the overdraft charging however, is that on this measure, the average level of liquidity structure, as fees are based on the outstanding overdraft at the recycling within each minute is considerably higher in CHAPS end of each minute. Provided that all payments are offset by Sterling than in Fedwire. In other words, the level of liquidity incoming payments within that same minute, irrespective recycling, whether as a result of active coordination or of the ordering of the payments, no charge is incurred. otherwise, appears to be greater throughout the day in CHAPS We choose to adopt the same methodology for the Sterling. This result is consistent with our hypothesis that the measurement of incoming payments as a funding source in CHAPS Sterling (Box 4). While recognising that payments cannot be “offset” within a minute in the same way as in Chart 8 Fedwire (if outgoing payments are made first, intraday Shares of Funding Sources of CHAPS Sterling liquidity will be required even if it is subsequently replenished System Payments by incoming funds), this measure is useful both as a point October 2006 of comparison with Fedwire and as an indicator of liquidity 18 Percent recycling in CHAPS Sterling. It should be noted, however, 100 that this particular measure does not capture the recycling Changes in account balances of funds hoarded from previous time periods, which is a 80 significant omission. We return to this point in a subsequent Throughput 60 discussion. CLS Bank pay-ins guidelines 7:00 11:00 12:00 14:30 40
20 18 It is also questionable whether active offsetting of payments within a minute Payments in is a realistic representation of members’ liquidity management processes. 0 However, as discussed in Section 2.5, the use of bilateral limits may result in 6:00 8:00 10:00 12:00 14:00 16:00 such behaviour being observed, since outgoing payments may be released Time immediately when incoming payments create headroom under a bilateral limit. Sources: CHAPS payment database; Bank of England calculations.
124 The Timing and Funding of CHAPS Sterling Payments Chart 9 Recycling and Liquidity Usage Shares of Funding Sources of Fedwire Daily Average, October 2006 Funds Transfers Average of Four Days over Half-Hour Intervals Liquidity Used (Minimum-Maximum Percent Recycling Ratio (r) Number of Banks Range, in Percent) 100 0 < r ≤ 5 2 68.0 - 85.1 Transfer of account balances 80 0 < r ≤ 10 5 37.0 - 99.7 r > 10 5 12.5 - 58.5 60 Sources: CHAPS payment database; Bank of England calculations. Extension of funds overdrafts 40 Notes: We exclude the Bank of England and CLS Bank. The Royal Bank of Scotland and Natwest are treated as a single entity (the Royal Bank of 20 Scotland Group), although they retain separate settlement accounts. Incoming transfers 0 from other banks 0:30 2:30 4:30 6:30 8:30 10:30 12:30 14:30 16:30 Time used (the recycling ratio). This measure is not subject to the Source: Federal Reserve Bank of New York. critique of the previous section, since it does capture the Note: Because few payments are made between 12:30 a.m. and 8:30 a.m., the variation in the shares of funding sources during that period of benefits of liquidity hoarded over multiple periods. the day is driven by a small number of payments. There is a wide variation in the extent of liquidity recycling achieved by CHAPS Sterling members (see table). Only five banks—all domestic—achieve recycling ratios greater than 10, and in each of these cases the proportion of liquidity drawn more concentrated structure of CHAPS Sterling is likely to be down is relatively low.19 The foreign banks achieve lower more conducive to liquidity recycling throughout the day, both recycling ratios and draw down a correspondingly large as a natural consequence of there being fewer participants and proportion of available liquidity. This implies that foreign as a result of bilateral coordination resulting from “small club” banks may indeed face greater liquidity constraints than behaviour. Such behaviour, in combination with the domestic ones, as suggested earlier in our discussion of the cost distribution of time-critical payments and the effect of of collateral. As noted, we would expect the incentive to delay throughput guidelines, may ensure that liquidity continues payments in order to take advantage of liquidity recycling to to flow smoothly through the system. be correspondingly high, but this expectation is not supported by the data. How might this be explained? From Section 3, we know that the intraday value profiles of outgoing payments made by domestic and foreign banks are 4.3 Liquidity Recycling and Liquidity similar. However, the ability to recycle liquidity and thereby Constraints lower aggregate liquidity requirements during the day also depends on the distribution of incoming payment flows. We noted earlier that the opportunity cost of posting collateral Charts 10 and 11 clearly illustrate that the pattern of net (or holding positive reserve account balances) may not be payments is very different on aggregate for domestic and uniform across all CHAPS Sterling participants. In particular, foreign banks, even though all members must comply with the cost of collateral, and hence of liquidity, should be higher the throughput guidelines. for foreign banks, since they are not subject to the Stock Domestic banks are, on aggregate, net recipients of funds Liquidity Regime. If this is the case, foreign banks would have in the morning and net suppliers of funds in the afternoon. a stronger incentive to fund payments using incoming funds, Foreign banks exhibit the opposite trend: Net payments are and they would be seen to attain higher recycling ratios. Is this negative until late morning and become positive thereafter. indeed the case? This implies that domestic banks tend to accumulate funds To answer that question, we consider the relationship during the morning and then pay these funds out in the between the maximum proportion of liquidity drawn down afternoon, thereby reducing intraday liquidity usage and and the overall level of liquidity recycling achieved by each increasing the recycling ratio. When the flows are reversed, member during the day, measured as the ratio of the value of total daily gross payments to the maximum value of liquidity 19 Note that foreign and domestic banks are not differentiated in the table.
FRBNY Economic Policy Review / September 2008 125 Chart 10 Chart 12 Net Payments for Domestic Banks in CHAPS Liquidity Usage by Domestic Banks Sterling System Daily Average, October 2006 Daily Average, October 2006 Percent Billions of British pounds 100 0.6 Maximum 80 0.4
0.2 60
0 40 Mean -0.2 20
-0.4 Minimum 0 -0.6 6:00 8:00 10:00 12:00 14:00 16:00 Time 6:00 8:00 10:00 12:00 14:00 16:00 Time Sources: CHAPS payment database; Bank of England calculations.
Sources: CHAPS payment database; Bank of England calculations.
Chart 11 Chart 13 Net Payments for Foreign Banks in CHAPS Liquidity Usage by Foreign Banks Sterling System Daily Average, October 2006 Daily Average, October 2006 Percent 100 Billions of British pounds Maximum 0.6 80 0.4 Mean
0.2 60
0.0 40 Minimum -0.2 20
-0.4 0 -0.6 6:00 8:00 10:00 12:00 14:00 16:00 Time 6:00 8:00 10:00 12:00 14:00 16:00 Time Sources: CHAPS payment database; Bank of England calculations.
Sources: CHAPS payment database; Bank of England calculations.
as is the case for foreign banks, liquidity must be drawn down While it is clear that the maximum proportion of liquidity to fund the net outgoing payments; hence, one would expect used by foreign banks is, on average, higher for much of the day recycling ratios to be lower. However, while the distinction than it is for domestic banks, there is considerable variation between the payment flows of the domestic and foreign banks within both sets of banks. Some domestic banks are net payers at the aggregate level is striking, these results conceal for much of the morning and use a large proportion of their considerable variation within both sets of banks. This is available liquidity early in the day. For these banks, the apparent from the intraday patterns of liquidity usage, recycling ratio is relatively low. Other domestic banks, by illustrated in Charts 12 and 13. contrast, use little or no liquidity for much of the day. These
126 The Timing and Funding of CHAPS Sterling Payments banks are more able to take advantage of funds received in the coordination as those in Fedwire, the observed degree of morning and achieve the highest recycling ratios of all CHAPS liquidity recycling appears to be high. We have also seen that Sterling members. By contrast, all foreign banks in CHAPS the intraday profile of payments is comparatively smooth. Sterling are net payers during the morning and approach their Taken together, these observations reveal that even if collateral maximum liquidity usage early in the day. posting is perceived to be costly by some banks—and hence a So while liquidity recycling appears to be relatively efficient “liquidity incentive to delay” does exist in CHAPS Sterling— in CHAPS Sterling, the extent to which individual banks other features of the system help avoid a prisoner’s dilemma benefit from recycling varies considerably. For those banks that equilibrium in which the majority of payments are delayed are net payers early in the day—including all of the foreign until late in the day. This serves to reduce the maximum banks and some large domestic banks—recycling ratios are liquidity required to make a given set of payments and hence much lower. One could argue that this results from the low cost the aggregate value of collateral that needs to be posted. The of liquidity, since the incentive to structure payments so as to empirical evidence suggests that payment coordination may also play an important role in Fedwire, although in this case coordination—and consequently liquidity recycling—is While liquidity recycling appears to be strongly concentrated around an end-of-day focal point. Which features of the system support this high and constant relatively efficient in CHAPS Sterling, the level of liquidity recycling? Centralised coordination devices extent to which individual banks benefit are likely to play a role; in particular, throughput guidelines from recycling varies considerably. counteract any generalised tendency to delay payments until the end of the day. Indeed, the spike in the proportion of payments “offset” before noon is evidence that the incoming reduce liquidity costs may be weak. But many of the banks with funds become an increasingly significant funding source at this relatively low recycling ratios do appear to face liquidity time of day, reducing the liquidity cost of complying with constraints, since they also use a large proportion of liquidity the deadline. posted. This suggests that some banks may be unable to recycle Other forms of “decentralised coordination” between liquidity to the extent that they would wish, which may reflect members may also be significant. The high visibility of the simple observation that coordination will result only if all payment flows in the concentrated CHAPS system allows (or a sufficient number) of the banks are similarly incentivised members to monitor their bilateral positions and to take action by liquidity pressures to cooperate. if counterparts fail to make payments in a timely fashion. The variation in recycling ratios also reflects the effect of the The prisoner’s dilemma may then simply be resolved through other influences on payment timing. The observed patterns of the repeated interaction of the small number of participants, whereby recalcitrant participants are “punished” for failing net payments in Charts 10 and 11 are likely to reflect structural to provide liquidity to the system. It is arguable that these differences in the underlying businesses of the participants and pressures may be less strong in the more diffuse Fedwire their customers, resulting in differences in the distribution of system. While not explicitly revealed by the aggregate data, payment instructions and deadlines. If, for example, certain there is also anecdotal evidence that participants apply bilateral participants (or their customers) routinely borrow in the net sender limits with respect to other system participants, overnight market while others lend, the payment flows of the thereby promoting the recycling of liquidity between each two groups will be correspondingly different. To the extent that bilateral pair and enhancing the liquidity efficiency of the these structural factors limit participants’ discretion over the system. An empirical question remains as to how often these timing of payments, this may explain the observed variation limits “bite” in practice, but a “small club” like CHAPS is a in the distribution of recycling benefits.20 natural environment for the application of such devices, which help generate a smooth payment profile. The patterns in the timing and funding of CHAPS Sterling payments described in this article would appear to be risk- 5. Discussion and Conclusions beneficial, for individual participants and for the system as a whole. The low opportunity cost of posting collateral and Our analysis indicates that even though the intraday liquidity the tendency for all members (domestic and foreign) to post regime supporting CHAPS Sterling payments does not give collateral at the beginning of the day help ensure that time- rise to the same incentives for minute-by-minute payment critical payments do not fail for want of liquidity. Moreover, 20 The patterns of funding flows in the overnight market are the subject the apparently low opportunity cost of collateral results in of ongoing research at the Bank of England. many banks providing a liquidity cushion in excess of that
FRBNY Economic Policy Review / September 2008 127 required to make time-critical payments. This not only some implications for the efficiency of the system itself. increases the resilience of that participant to liquidity shocks, However, we have not been able to disentangle fully the effects but also contributes to the resilience of the system as a whole. of the factors identified; this is perhaps an inevitable drawback The efficient recycling of liquidity during the day further given our descriptive approach. Formal analysis, supported contributes to a reduction in liquidity risk by reducing the and complemented by further econometric work on payment aggregate liquidity required to make a given set of payments. data, may help in this direction by shedding additional light on In theory, this is particularly beneficial for those banks that face key issues such as the effect of membership size on payment a relatively high cost of liquidity. However, we have seen that behaviour and the precise way in which banks achieve many of the members that draw down a large proportion of coordination (using, for example, bilateral net sender limits). available liquidity are unable to take advantage of incoming Our analysis of the variations in liquidity recycling intensity funds to the same extent as other members, perhaps owing to also makes a strong case for further analysis of the overnight the distribution of underlying payment orders from customers. market, particularly its effect on payment behaviour and For these banks, the likelihood of needing to post additional liquidity usage. Ultimately, such analysis will contribute to collateral during the day may be correspondingly high. an understanding of how a large-value payments system Our analysis can take several interesting directions. In this functions and suggest where and how risks to the system article, we have formulated a number of hypotheses on the may crystallise. determinants of behaviour in a payments system, suggesting
128 The Timing and Funding of CHAPS Sterling Payments Appendix: Effects of a Bilateral Net Sender Limit on a Bank’s Behaviour
δ()> > δ()< 21 We present a simple, formal illustration of a single bank’s Qt 0 if Q 0 , and Qt 0 otherwise. To < problem when choosing an optimal level of liquidity. A simplify further, we assume that if Qt 0 , the bank sells bilateral net sender limit is shown to incentivise early liquidity the spare liquidity in the market, immediately realising δ() > δ() provision and early payment, thereby generating high levels Qt ; if instead Qt 0 , then Qt includes all costs of liquidity recycling. No attempt is made here to develop a derived from delaying payments, in particular, the cost fully fledged game-theoretical model of payments. That model of the extra liquidity with which the bank settles or would be more complex, because incoming payments would cancels the queued payments. need to be modeled as a strategic choice (by the other banks) • The bank then begins the next period t + 1 afresh, with instead of as an exogenous random variable. Considering a no liquidity and no queues (all costs / benefits stemming δ() single bank’s decisions in isolation allows us to focus on the from Qt are accounted for by Qt ). marginal effect of a bilateral net sender limit on the incentives We now look at the bank’s incentives and its optimal choice. to post liquidity and to delay payments. We would nevertheless We want to show how a bilateral net sender limit incentivises expect to find this effect in a more complex setting. short queues and thus liquidity recycling. To do so, we compare two cases, one in which there is no bilateral net sender limit and one in which a limit is implemented.
Setting
Suppose that our bank receives payment orders exogenously The Bank’s Problem: I from its customers. To execute these orders, the bank requires liquidity, which can be obtained either by posting collateral or by waiting for exogenous (and random) incoming payments. We assume that our bank faces the following sequence of No Bilateral Sender Limits events, all of which occur within a fixed time interval t (which can be thought of as a metaphor for a trading day, or part of Suppose that incoming payments yt arrive according to some the day such as “the morning”): exogenous distribution f (.), which is independent of t and of the bank’s choices. In this case, the bank’s problem is 22 t.0 ______t.1______t.2______t.3 actually a single-period maximisation, so we are able to , δ() eliminate all time indices. By borrowing w, the bank reduces xt wt yt pt Qt . the expected queue Q , thus abating the expected delay costs δ()Q (and possibly transforming them into a gain, if Q < 0 ). •At t.0, the bank receives payment orders to the value xt . However, liquidity is costly, and the bank may hope to make •At t.1, the bank decides how much liquidity to raise, use of receipts from the other bank (via y). In general, the bank w , at a cost λ()w . t t will not raise a full wx= , and it will rely partly on incoming •At t.2, incoming payments provide the bank with payments. additional liquidity yt , so the bank has total liquidity of lt = wt + yt . 21 A negative queue is a positive amount of liquidity whose positive value p l < x is a negative cost. Note that, to rule out the existence of “money-making •At t.3, the bank makes payments t . If t t , the bank δ λ () machines,” and (the cost of liquidity) are mutually restrained. See the can only pay up to lt pt = lt , so it “queues” an amount example below. > of payments Qt = xt – lt . If instead lt xt , then pt = x t 22 There are no spillover effects between t and t + 1 because we assume that δ() and the bank has spare liquidity. To simplify, we assume the bank realises Qt from any queue or spare liquidity, beginning δ() that the cost of a backlog Qt is a function Qt , with period t + 1 afresh.
FRBNY Economic Policy Review / September 2008 129 Appendix: Effects of a Bilateral Net Sender Limit on a Bank’s Behaviour (Continued)
To find the optimal w, the bank looks at the total cost The Bank’s Problem: II (Example) from raising w : We now solve analytically the bank’s problem, under particular Cw()= λ ()w + Eδ ()Q assumptions about the cost functions and the distribution of y. r =fyλ ()w + ∫ ()δ()xw– – y dy , Suppose the delay cost is 0 ⎧ CQ if Q ≥ 0 where r is an upper bound on the incoming payments that the δ ()Q = ⎨ , ⎩ cQ if Q < 0 bank can expect to receive. Provided that some technical convexity conditions hold, the optimal amount of w∗ is with 0 < Bilateral Net Sender Limit When a bilateral net sender limit is in place, the distribution of Q incoming payments y is no longer exogenous, but depends on previous payments. As a trivial example, if too few payments are sent out, the bilateral limit is hit and no further incoming y In this case, the costs of a positive queue Q grow faster than the can be expected. More generally, the effect of choosing a gains from spare liquidity (which are simply the negative costs particular w , and hence a volume of payments p , spills from a negative queue). t – 1 t – 1 λ() λ over to the next period and influences the expected amount of We also assume that liquidity costs are linear: w = w λ > incoming funds y . As a consequence, compared with the case with 0 . To make the problem interesting, we first impose t λ < in which no bilateral net sender limits exist, every unit of C, implying that it is better to post liquidity and make a payment than not to post it and fail to make the payment, and liquidity made at t –1 now carries an extra benefit in terms of a then impose c <λ , which ensures that it is not optimal to post liquidity saving at time t (although, of course, only in expected infinite liquidity. Indeed, if it were c >λ , then any pound of terms and only if it actually allows additional payments to be liquidity would be worth more to the bank than the cost λ , made at t –1 ). This interperiod spillover tilts the balance in independently of whether it is used to shorten a queue (which favour of posting more w , increasing payments and reducing t gives a benefit C > λ ) or if it results in spare liquidity (whose the queue. If, in addition, the liquidity cost λ increases in time, benefit is precisely c, which must therefore be assumed to be this mechanism is reinforced: The liquidity saved at t +1 by smaller than λ ). posting more wt at t is even more valuable. Similarly, if the bilateral limit itself depends on i’s past payments (for example, a “bad” payment record may induce the other banks to tighten prudentially their bilateral limit toward i ), the incentives to pay early will be even stronger. 130 The Timing and Funding of CHAPS Sterling Payments Appendix: Effects of a Bilateral Net Sender Limit on a Bank’s Behaviour (Continued) No Bilateral Net Sender Limit 1 Suppose y is uniformly distributed in []0, r , so its probability density function is fy()= --- . t r Then, the bank’s cost function is: r (1) Cw()= λ ()w + ∫ f ()y δ ()xwy– – dy = o 0 1 xw– r λ w + --- Cxwy∫ ()– – dy+ c∫ () x– w– y dy if xw– ≤ r ⎧ r xw = ⎨ 0 – ⎩ 1 r λw + --- Cxwy∫ ()– – dy if rxw< – r 0 1 1 2 1 2 λ w + --- C---()wx– – c---()rwx+ – if xwr– ≤ ⎧ r 2 2 = ⎨ . ⎩ 1 w()λ – C – ---Cr()– 2x if rxw< – 2 We now find the optimal w, to be called w∗ . Finally, substitution of equation 2 into equation 1 yields the optimal (minimised) cost: • Suppose w∗ is such that xw– ∗ ≤ r . In this case, the 1 2 optimality condition would be (3) C ∗ = ------()–rλ ++2Cxλ 2crλ – 2cxλ – Ccr . 2()Cc– d ⎛⎞1 1 2 1 2 (Opt) ------λw + --- C---()wx– – c---()rwx+ – 1 dw⎝⎠r 2 2 If we set, for example, c = --- , λ ===rx1 , and C = 2 , 1 () 2 ==---()rλ + Cw– Cx– cr– cw + cx 0 the graph of Cw is: r rc()– λ yielding an optimal liquidity w∗ = x + ------and thus C(w) 1.2 Cc– a total cost equal to 1.0 1 2 ------()– rλ ++2Cxλ 2crλ – 2cxλ – Ccr = C ∗. 2()Cc– 0.8 < ∗ λ < • Suppose instead rxw– . Because – C 0 , the total 0 0.5 1.0 1.5 W cost decreases in w as long as rxw< – (see equation 1). ∗ Hence, we would have a corner solution at w = xr– , In this case, the liquidity posted is 66 percent of the 1 which yields a cost xλ – rλ + ---Cr = C ∗∗ . payments due (x = 1 ). 2 1 r 2 Now, the difference C ∗∗– C∗= ------()C – λ is always 2 Cc– positive. Hence, the cost-minimizing w is the one found in the Bilateral Net Sender Limit (BNSL) first case, supposing xw– ∗ ≤ r : rc()– λ Suppose again that yt is uniformly distributed. This time, (2) w∗ = x + ------< x . Cc– however, imagine that incoming payments are drawn from []0, r , with r determined by a bilateral net sender limit: It should be noted that w∗ has the anticipated properties: t t It increases in x (the amount of payments to make) and in C () rt = rt – 1 + pt – 1 – yt – 1 . (the cost of queues), and it falls with λ – c , that is, with the difference between the cost of liquidity and its benefits as In this case, an increase in pt – 1 due to higher liquidity end-of-day spare liquidity. wt – 1 pushes up rt , which in turn affects the minimised costs FRBNY Economic Policy Review / September 2008 131 Appendix: Effects of a Bilateral Net Sender Limit on a Bank’s Behaviour (Continued) (see equation 3). What is the value of such a spillover, from where the last inequality comes from equation 2. Combining t – 1 liquidity into time t costs? The answer is the equation above and equation 4, we finally have ∗ ∗ dC dC dr ∗ ------t = E ------t------t dC t – 1 , ------t = ZW < 0 dwt – 1 drt dwt – 1 . dwt – 1 which we calculate term by term. As anticipated, the spillover is negative, corresponding to an ∗ dC abatement of costs. When the bank internalises these spillovers, The term ------t- immediately derives from equation 3, which drt these gains (perhaps discounted by a factor β ) are added to the reveals that the expectation term is irrelevant here (λ , c, and C optimality condition (Opt), which therefore becomes are constant over time): ∗ dCt – 1 dCt 1 ∗ ------+ β------= ---()rλ + Cw– Cx– cr– cw+ cx – ZW = 0. dC 1 2 dw dw r (4) ------t- = – ------()λ + Cc – 2cλ = Z < 0 . t – 1 t – 1 dr 2()Cc– t Hence, the BNSL shifts down the marginal cost schedule. To prove the last inequality, recall that c <<λ C . This This clearly brings about a higher level of liquidity posting in implies both λc < λ2 and λcCc< , which, summed member t – 1 (in the example below, from 66 percent to 80 percent). by member, gives 2λc < λ2 + Cc . Total and marginal cost of w, with (blue) and without dr Expectations do matter on the second term ------t . In fact, (black) a BNSL: dwt – 1 w dC/dw an extra pound worth of t – 1 translates into one more 1.0 payment only if the available liquidity wt – 1 + yt – 1 turns out to C 0.5 be less than the payment orders xt – 1 , with yt – 1 unknown at t – 1. Formally, 1.0 0 0.8 ⎧ < -0.5 drt 1 if wt – 1 + yt – 1 xt – 1 ------= ⎨ . 0.6 dwt – 1 ⎩0 otherwise -1.0 0 0.5 1.0 1.5 0.5 1.0 1.5 Hence, the expectation is calculated as: W W dr xt – 1 – wt – 1 t () 1 ()> Et – 1 ------= ∫ 1fsds = ------xt – 1 – wt – 1 = W 0, dwt – 1 0 rt – 1 132 The Timing and Funding of CHAPS Sterling Payments References Armantier, O., J. Arnold, and J. McAndrews. 2008. “Changes in the Kobayakawa, S. 1997. “The Comparative Analysis of Settlement Timing Distribution of Fedwire Funds Transfers.” Federal Reserve Systems.” Centre for Economic Policy Research Discussion Bank of New York Economic Policy Review 14, no. 2 (September): Paper Series, no. 1667, July. 83-112. Martin, A., and J. McAndrews. 2008. “Liquidity-Saving Mechanisms.” Bech, M. L., and R. Garratt. 2003. “The Intraday Liquidity Journal of Monetary Economics 55, no. 3 (April): 554-67. Management Game.” Journal of Economic Theory 109, no. 2 (April): 198-219. McAndrews, J., and S. Rajan. 2000. “The Timing and Funding of Fedwire Funds Transfers.” Federal Reserve Bank of New York Becher, C., S. Millard, and K. Soramäki. 2007. “The Network Topology Economic Policy Review 6, no. 2 (July): 17-32. of CHAPS Sterling Payments.” Unpublished paper. Mills, D. C., and T. D. Nesmith. 2008. “Risk and Concentration Beyeler, W., R. Glass, M. Bech, and K. Soramäki. 2006. “Congestion and in Payment and Securities Settlement Systems.” Journal Cascades in Payment Systems.” Federal Reserve Bank of New York of Monetary Economics 55, no. 3 (April): 542-53. Staff Reports, no. 259, September. Soramäki, K., M. L. Bech, J. Arnold, R. J. Glass, and W. E. Beyeler. 2006. Buckle, S., and E. Campbell. 2003. “Settlement Bank Behaviour and “The Topology of Interbank Payment Flows.” Federal Reserve Throughput Rules in an RTGS Payment System with Collateralised Bank of New York Staff Reports, no. 243, March. Intraday Credit.” Bank of England Working Paper no. 209. James, K., and M. Willison. 2004. “Collateral Posting Decisions in CHAPS Sterling.” Bank of England Financial Stability Review 17, December: 99-104. FRBNY Economic Policy Review / September 2008 133 Risk Management in Payments Systems Articles by: Michele Braun, James McAndrews, William Roberds, and Richard Sullivan Antoine Martin and David C. Mills Michele Braun, James McAndrews, William Roberds, and Richard Sullivan Understanding Risk Management in Emerging Retail Payments • The retail payment landscape is shifting 1.Introduction increasingly from paper to electronic form as the number of ways to make noncash lectronic checks, cell phones, and speed-through lanes payments grows. E at toll booths are just a few examples of new payment methods recently introduced to the market. Based on • Payment products, services, rules, and computer technology, online commerce, and telecom- technologies are changing at a rapid rate— munications, these new payment methods rely on electronics as are the tools for perpetrating fraud, illicit for most or all of their functions. Many products based on use, and breaches of data security. these methods have failed, some have struggled to grow, and a few have become well accepted in routine commerce. All face • Providers of emerging payment methods now a variety of risks. Reflecting these risks, news reports of data breaches, identity face the same risks as providers of more theft, and fraud have become a part of the electronic payment established methods; failure to control these landscape. Novel characteristics associated with “emerging” risks can lead to rejection in the market. payments include low-cost ways to store and transmit data. These technologies can reduce risk, but they can also lead to • By limiting access to payment networks, new risks. It is timely now to develop a structure and monitoring for compliance with risk mitigation vocabulary for examining how new payment technologies standards, and enforcing penalties for affect risk, particularly as the number of ways to make noncash noncompliance, emerging as well as payments grows and as payments shift from paper-based to established providers can contain many of electronic form.1 the risks associated with fraud, illicit use, Understanding the structure of risk is useful, although and data security breaches. assessing losses and mitigation efforts in a new payment Michele Braun is an officer and James McAndrews a senior vice president at This article began as a Federal Reserve System staff research project. the Federal Reserve Bank of New York; William Roberds is an economist and The authors acknowledge those who participated in that research, each policy advisor at the Federal Reserve Bank of Atlanta; Richard Sullivan is a contributing significantly to the concepts and substance of this article: Eugene senior economist at the Federal Reserve Bank of Kansas City. Amromin, Peter Burns, Marianne Crowe, Jim Cunha, Dan Littman, Mark Correspondence: FRBNY Economic Policy Review / September 2008 137 product can be difficult. Low levels of fraud losses, for example, 2. True Accounts of Fraud and could imply that: 1) risk is low, 2) current mitigation practices Operational Risks in Payment are effective, or 3) weaknesses have not yet been discovered. Innovations However, high levels of losses demonstrate that risks are high, and it takes time to know whether mitigation efforts can The following accounts illustrate fraud and operational succeed. In either case, only time and the monitoring of problems that exploited the novel characteristics of new problems will reveal whether risk can be controlled sufficiently. payment methods. These incidents include a telemarketing In this article, we consider whether, in this period of scheme, a complex online fraud, and two data security uncertainty, the sponsor of an emerging payment method has breaches. The crimes that underlie these incidents—fraud, enough incentives and tools to control risk before the harm con artistry, and theft of money, property, or someone’s good from fraud or operational problems becomes widespread.1 name—are not themselves new. The operational problems are Our analysis suggests that the sponsors and providers of also not necessarily new, but the potential scale and speed of successful emerging payment methods must be aware of the disruptions are of a magnitude untypical of their paper- potential fraud risk and operational risk. Moreover, they must based counterparts. It is timely now to develop a structure and 2.1 Telemarketing Fraud vocabulary for examining how new payment technologies affect risk, particularly as In 2003, the Federal Trade Commission (FTC) announced that the number of ways to make noncash it had closed down the Assail Telemarketing Network and its affiliates. The FTC alleged that the Assail companies ran payments grows and as payments shift telemarketing activities from so-called boiler-room operations from paper-based to electronic form. that offered credit cards to consumers with poor credit records.2 Under the guise of charging membership fees, these firms persuaded consumers to provide the bank and account information from their checks.3 The telemarketers then used mitigate these risks or face rejection in the payment market. this information to create electronic debits to consumers’ Service providers can contain risks by limiting access to their checking accounts as payment for the “membership” fees. payment networks, monitoring for compliance with risk These credit cards appear to have been rarely, if ever, delivered. mitigation standards, and enforcing penalties for The consumers found, however, that they had also been signed noncompliance. While much of this containment activity is up for expensive and dubious products (so-called upsell voluntary, some is enforced by public authorities that can help programs) such as auto club memberships, the fees for which coordinate activities as well as define and enforce standards. were directly charged to their bank accounts. When consumers This article explores in several ways the structure and called to complain, the companies used elaborate scripts to vocabulary of emerging payments system risks and their avoid repayment or cancellation of the membership. The FTC mitigation. We begin by recounting several incidents of fraud alleged that Assail and its principals engaged in deceptive and losses associated with emerging payment methods. We marketing activities that totaled more than $100 million.4 then describe an economic framework for understanding risk The particular type of electronic transaction that Assail control in retail payments. Next, we apply the framework to the used, a debit through the automated clearinghouse (ACH), risk experiences of three new payment types. These must be processed, collected, and paid through participating approaches—both deductive and inductive—are banks. These banks are supposed to monitor the companies for complementary ways to understand risk and its mitigation in emerging payment methods. Finally, we discuss some general 2 See Federal Trade Commission, “International Telemarketing Network observations derived from integrating the economic concepts Defendants Banned from Telemarketing,” press release, January 24, 2005, and actual experiences, then offer conclusions. available at 138 Understanding Risk Management in Emerging Retail Payments which they provide this ACH origination service. In this case, This was a case of fraudsters hacking into databases, stealing First Premier Bank admitted that it had failed to perform due payment-related and other information, using the stolen diligence on the activities and legitimacy of its customers, but identities to create fictitious accounts, manipulating online it then helped identify the telemarketers and supplied auctions, and using machine-based tools to proliferate their information to the investigative agencies. The bank later paid thefts and confound the transaction/audit trail. $200,000 to Iowa, South Dakota, and Minnesota as part of a Ultimately, the FBI used an undercover operation to lure the wider settlement and agreed to engage vigorously in know- two hackers to Seattle, Washington, where they had been your-customer practices and ongoing monitoring of customer invited under the pretext of a job interview with “Invita,” a activity.5 fictitious computer security company. In October 2002, the Before the particular ACH transaction type used by Assail two men were sentenced to three years in prison. was introduced, this type of fraud was often perpetrated by creating a “remotely created check”—a check that contains a text legend in lieu of the payer’s signature. This approach is still used to commit fraud, but it does not offer the speed and scale 2.3 Unsecure Data this fraudster achieved using automation.6 In 2005, the president and chief executive officer of CardSystems Solutions, Inc., a transaction processor, testified before a Congressional committee that, in September 2004, 2.2 Transaction Fraud and Data Security an unauthorized party had placed a clandestine computer Breach program on the company’s transaction processing system (Perry 2005). CardSystems reported that, on May 22, 2005, it The U.S. Department of Justice reported that, in 2000, two suffered a “potential security incident.” Records on 263,000 Russian men, Vasiliy Gorshkov and Alexey Ivanov, used transactions were stolen—including account holders’ names, unauthorized access to Internet service providers in the United account numbers, expiration dates, and security codes. Forty States to misappropriate credit card, bank account, and other million records were potentially at risk. personal financial information from more than 50,000 CardSystems disclosed the breach to its bank as well as to 7 individuals. They allegedly hijacked computer networks and MasterCard, Visa, and American Express. The three credit card then used the compromised processors to commit fraud companies determined that CardSystems had violated the through PayPal and the online auction company eBay. credit card industry’s prevailing security and data retention According to the Justice Department’s press releases, the standards. Visa and American Express announced that they fraudsters developed elaborate programs to establish would not permit the firm to process their transactions after thousands of anonymous e-mail accounts at websites that, at October 31, 2005. On October 15, Pay by Touch announced its the time, did not have the sophisticated tools required to acquisition of CardSystems Solutions because of the latter’s distinguish human intervention at set-up. Gorshkov’s network connections to 120,000 merchants, despite the demise programs created accounts at PayPal that were based on of its card transaction processing business.9 random identities and stolen credit card numbers. The More recently, in early 2007, the TJX Companies, which programs then transferred funds from one account to another operate retail stores in the United States, Canada, Ireland, and to generate cash and to pay for computer parts purchased from the United Kingdom, reported that data security breaches from vendors in the United States. Additional computer programs mid-2005 until late 2006 might have compromised more than allowed the conspirators to control and manipulate eBay 45 million customer records.10 Company investigations also auctions so that they could act as both seller and winning revealed breaches in 2003 and 2004, as well as compromised bidder in the same auction and then effectively pay themselves driver’s license numbers and addresses. The Massachusetts using the stolen credit cards.8 Bankers Association reported fraudulent use of debit and credit 5 This was the first time that the Federal Trade Commission tried to hold a bank cards issued by its members as a result of that breach. The responsible for the deceptive practices of its customer. 8 6 To help reduce the potential for fraud in the use of remotely created checks, Physor.com describes some of the techniques used by criminals to perpetrate the Federal Reserve Board amended its Regulation CC effective on July 1, 2006, fraud through online auction sites. See FRBNY Economic Policy Review / September 2008 139 Association’s press releases recounted that fraudulent card data methods that differ from established payments in a significant had been used to make purchases in many U.S. states, Hong way—that is, technologically, contractually, legally, or Kong, Sweden, and other countries.11 conceptually. The Wall Street Journal reported that hackers first tapped Third, every payment method involves risk. The Bank for into data transmissions from handheld equipment used International Settlements’ Committee on Payment and to manage store inventory and prices.12 Reportedly, they Settlement Systems identifies five major categories of risk used these captured data to crack encryption codes and to associated with payment transactions: fraud, operational, legal, steal employees’ user names and passwords at company settlement, and systemic.13 Generally, other types of risk are headquarters. With the resulting access to TJX’s network, subcategories of these five broad types. Emerging payment they stole credit and debit card numbers and even left messages methods may be particularly susceptible to fraud and for each other. Stolen card numbers were then allegedly sold on operational risks. They may also carry enhanced legal risk the Internet. Press reports traced losses to banks across the simply because case law is less well developed or because the country. In addition to direct purchases with stolen credit drafters of established laws and regulations may not have and debit card numbers, the thieves or their customers also foreseen some of the ways in which payments are initiated, purchased prepaid cards, which were in turn used to purchase processed, and settled. Definitions of the three risks mainly goods and services. associated with emerging payments are presented in the box. A payment method may also carry risks not directly associated with the success or failure to transfer value. Instead, indirect problems may arise that appear ancillary to the 3. Definitions and Economic Insights financial transaction. For emerging retail payment methods, two risks of this type are notable: data security risk and risk of The examples just offered illustrate some risks of financial loss illicit use. In these cases, the payment methods function and that are present in payment methods. We now turn to an transfer value correctly, but something underlying the economic examination of these risks and their mitigation, transaction is “bad.” beginning with three general observations. First, the risks Data security risk is a form of operational risk involving present when new or still-emerging payment methods are used unauthorized modification, destruction, or disclosure of data are not wholly different from those present in long-established used in or to support transactions. For example, a data security methods of payment. Nonetheless, our analysis suggests that certain risks are more salient in emerging retail payments than elsewhere in the payment marketplace. The risks present when new or still- Second, new payment methods are generally based on, or emerge from, existing payment products. To focus this emerging payment methods are used are discussion, we define established payments to include paper not wholly different from those present in checks, recurring transactions transferred through the ACH, long-established methods of payment. credit card and debit card transactions made with magnetic- stripe cards, and wire transfers. To this base, enhancements, innovations, and rules are added to address newly identified breach may facilitate identity theft, which could trigger later market opportunities or to take advantage of expanding harm to a party in a transaction or an otherwise uninvolved technical capabilities. Sometimes innovations are sufficient to party elsewhere in the system. yield a distinguishably new payment method. Thus, we define Risk of illicit use is the risk that a payment method may be emerging retail payments as those newly introduced payment used for illegal purposes, for example, money laundering, 11 Massachusetts Bankers Association, “Massachusetts Banks Now Reporting terrorism financing, or the purchase of illegal goods and That Fraud Has Occurred Due to the TJX Data Breach,” press release, January 24, services such as drugs or child pornography. Similarly, the ease 2007, available at 140 Understanding Risk Management in Emerging Retail Payments Major Risks in Emerging Payments nonrival goods are recorded music, video, and computer software. The data managed by modern payments systems are Type of Risk Definition another example of this type of good: The use of a credit card in one electronic transaction does not diminish the ability to Fraud Risk of financial loss for one of the parties involved in a payment transaction arising from wrongful or criminal deception. The risk that a transaction cannot be properly completed Any information stored and transmitted by because the payee does not have a legitimate claim on the payer. a payments system can be described as Operational Risk of financial loss due to various types an economic good, an item having value of human or technical errors that disrupt the clearing and settlement of a payment in exchange. transaction. The risk that a transaction cannot be properly completed due to a defective device or process that precludes the completion of all the steps required in a transaction. use it in another transaction so long as the credit limit is not Legal Risk that arises if the rights and obligations exceeded. (Credit, cornflakes, and sneakers are not nonrival of parties involved in a payment are subject goods; they get used up.) to considerable uncertainty. Central to the value of any digital good is data integrity— garbled music or video is useless, for example. The usefulness Source: Bank for International Settlements (2000). of payment data can be diminished by fraud and security breaches or by operational disruptions that make it difficult to transmit data. Consequently, we argue that the integrity of payment data is also a nonrival good. If a payments system participant secures a facility against operational disruptions 3.1 Some Insights from Economic Theory and fraud, it creates an environment conducive to smooth operation of the payments system, generating benefits for other participants as well. Nonrival goods are classified as club goods or public goods Risk Containment as a Good according to whether access to the good can be limited. A club good is a nonrival good that a group or individual can be Economic theory offers some useful concepts for under- stopped or excluded from consuming. For example, cable standing risk in payments systems. All payments systems are television firms exclude nonsubscribers from their service by systems for managing valuable information: They keep records encoding their signals and giving decoders only to paying of transactions and communicate transaction data. Any subscribers. A public good is a nonrival good for which access information stored and transmitted by a payments system can cannot be limited. National defense, for example, is a nonrival be described as an economic good, an item having value in public good because everyone in a country is covered and no exchange. one can be excluded from the benefits. Thanks to modern information technology, emerging In the case of actions to contain fraud and operational risks payment methods can offer tremendous efficiency gains over in emerging payments, the club good description is perhaps the traditional methods of making payments. Electronic data can most appropriate. Successful private sector payment providers be easily stored at a few locations and then shared among (for example, credit cards, debit cards, and ATM networks) payments system participants at very low cost. Payment data have by and large managed to contain fraud.14 They also thus meet Varian’s (1998) description of a digital good, a good maintain operating procedures and auditable controls to limit that can be stored and transferred in digital form. operational risk. Participation in these systems is limited by Varian argues that digital goods are different from standard, membership rules, and participants (individuals, merchants, physical goods (such as cornflakes, sneakers, and minivans) in that they are nonrival goods. A nonrival good is one whose value 14 Reported fraud rates for credit card transactions are about 5 basis points of does not diminish with any one individual’s use or value, and similar fraud rates are reported for checks (Nilson Report). Industry consumption of it. A textbook example of a nonrival good is representatives report that actual rates may be a little higher (Green Sheet). broadcast television: One’s consumption of a TV show does Visa reports an operational “reliability rate of 99.999 percent” (“Securing Payments: Building Robust Global Commerce,” 2005, available at FRBNY Economic Policy Review / September 2008 141 banks, and processors) associated with high levels of fraud or potential for rapid propagation of fraud and operational operational snafus can be expelled. disruptions is the flip side of the efficiency of electronic There are natural limits to the power of exclusion, however. payments.15 Since every payments system is a type of communications Varian (2004) points out some difficulties in the provision network, excluding too many network participants lowers the of such nonrival goods. Because the amount of mitigation network’s value for those parties that remain. There will always depends crucially on the participant that exerts the least effort, be a trade-off between security and inclusiveness. and because different system participants have different amounts at stake, there is a significant risk that participatory incentives will not be uniform. Participants with a lot at stake—that is, those with high net benefits from more Why Containing Fraud and Operational mitigation activity—will prefer a higher level of protection Risks Is Difficult from the risk in question than those with lower net benefits are willing to support. However, because overall protection Hirshleifer (1983) describes a model of a nonrival good that is depends on the participant that exerts the least effort, the latter particularly applicable to data integrity in electronic payments group determines the overall level of risk mitigation. systems. He describes the problem of people living in a The problem of nonuniform risk management incentives “polder,” a low-lying patch of land protected from flooding by crops up regularly in payment situations. Various stakeholders a system of dykes. Each resident of the polder is responsible for in payments systems will naturally prefer different levels of maintaining the portion of the dyke that abuts his or her mitigation in the system. The longer the supply chain or the property. The dyke clearly provides a nonrival communal larger the network for a given payment technology, the greater good: flood protection for all residents of the polder. is the potential for disagreement about the appropriate level of In this example, the degree of flood protection provided mitigation. depends exclusively on the height of the lowest portion of the Many different providers of services are integral to the dyke. In other words, the degree of protection will not be processing of electronic payments. These providers include determined by the total flood-mitigation efforts of everyone encryption firms, processors that route transaction data, and Internet service providers, among others. However, because A data breach or operational disruption in minimizing fraud and operational risks requires effort from all participants, some mechanism is needed to give all participants one portion of a payments system can the right incentives to “maintain the dyke.” Private contracts, open the metaphorical floodgates to laws, and regulations can each play a beneficial role in creating such incentives. problems throughout the entire system. living inside the dyke, but rather by the one resident who exerts Confronting Fraud and Operational Risks the least effort in maintaining the dyke. The analogy with emerging payments is straightforward: The risk mitigation Despite the difficulties outlined above, experience has shown effort of each party in the particular payments system to that all successful payments systems have learned to keep fraud maintain data integrity prevents fraudulent data from and operational risks at fairly low levels. Competition among circulating in the system, and a commitment to operational payments systems gives important incentives to service excellence allows others in the system to complete their providers to mitigate these risks. Systems that fail to contain transactions effectively. risks do not survive in the payment marketplace.16 There are obvious parallels between flood protection in Service providers have developed three broad approaches to Hirshleifer’s polder model and the mitigation of fraud and managing various kinds of payment risk: pricing, insurance, operational risks in payments systems. The 2005 data breach and containment. at CardSystems Solutions, for example, resulted in problems not only for CardSystems, but also for numerous other users 15 of card payments systems—cardholders, merchants, banks, For a formal exposition of this point, see Kahn and Roberds (2005). 16 and processors. A data breach or operational disruption in “Thinking Like a Criminal,” Arizona Republic, August 24, 2006, recounts how an entity that tried to compete with PayPal in the mid-1990s was closed one portion of a payments system can open the metaphorical down by Visa because as many as “three out of five . . . transactions turned out floodgates to problems throughout the entire system. The to be fraudulent.” 142 Understanding Risk Management in Emerging Retail Payments • Pricing means that a party that bears a risk is Of course, moral hazard can arise on the payer’s side, too— compensated appropriately. Pricing is extremely for example, when the right of a credit card holder to dispute a important in allocating credit risk—banks that issue transaction may tempt the cardholder to claim that fraud was credit cards charge higher prices, in the form of higher committed when it was not. Authentication procedures, interest rates on borrowing and higher annual fees for cards, to subprime cardholders who they believe are less likely to pay their balances. Issuing banks willingly bear a high level of credit risk on these cards because the higher Experience has shown that all successful interest earned compensates for the greater risk taken.17 payments systems have learned to keep • Insurance is an agreement between two parties as to who fraud and operational risks at fairly low will bear a loss when one occurs. Thus, for instance, a merchant that receives a credit card payment is insured levels. against the risk that the cardholder will not be able to pay the balance. particularly the collection of signatures at the point of sale, are • Containment is a catchall term for activities that tend to designed to contain this form of moral hazard. deter or suppress risk. In the case of fraud risk, examples Moral hazard can also lead to opportunistic behavior that include swiping a credit card through a card reader to magnifies operational risk in payments systems. A payment verify that the card is valid and asking for extra identification. processor might fail to spend the resources to maintain sufficient backup facilities—in the case of, say, a natural For fraud risk in particular, the effectiveness of the pricing disaster knocking out a key data center—because the negative and insurance approaches is limited by factors known as consequences of failing to maintain backup data do not accrue 18 adverse selection and moral hazard. fully to the processor, but rather to thousands of other Adverse selection refers to situations in which undesirable individuals and businesses as well. Card networks impose outcomes result from asymmetric information among various backup and resiliency standards to offset the lack of private parties to a transaction. Pricing works best to offset risks that incentives and contain this particular risk. are known and can be quantified in advance. When the payee But pricing and insurance alone are not sufficient risk and payer are anonymous to each other, the payee cannot management techniques: Credit card issuers do not seek out know if the payer poses a bad risk and is likely to make a cardholders who are likely to commit fraud, then attempt to fraudulent payment. Correspondingly, the payer cannot know recover the costs through differentially higher fees or interest if the payee is selling legitimate goods. Particularly when rates; ACH operators do not offer two fee schedules, one for commerce is conducted remotely (for example, over the reliable and another for unreliable originating banks; and Internet or by telephone), adverse selection undermines providers of payment services are generally reluctant to give incentives to play by the rules. “Bad actors” can optimize their unknown buyers and sellers guarantees against loss.19 own malign incentives, undermining the confidence of legitimate merchants and consumers. Moral hazard describes the effect of insurance on the incentives and thus behavior of an insured party. The Containment Techniques availability of insurance can lead to opportunistic behavior on the part of the insured at the expense of the insurer. For Containment of fraud and operational risks requires coopera- example, a merchant that accepts payments via cards branded tion among payments system participants. All need to have by a major network like Visa, MasterCard, or American incentives to undertake actions that will keep fraud and Express is insured against credit risk (and sometimes fraud operational risks down to acceptable levels. These incentives risk) and consequently may not have an incentive to make sure can be provided by monitoring system participants and then that a payment is legitimate and within a cardholder’s credit imposing penalties for inadequate risk controls that can lead limit. The card networks and their issuing banks, which provide to significant losses or disruptions. the insurance, contain the risks by imposing on merchants Monitoring is the foundation of containment: Checking on authorization and authentication procedures that create participants will reveal whether they are engaging in appro- appropriate incentives and guard against fraudulent card use. priate levels of risk mitigation. But monitoring is unlikely to be 17 About 4 percent of balances are never paid off. 19 Provisions in the Federal Reserve’s Regulations E and Z, which implement 18 These problems generally plague information security; see Anderson and the Electronic Funds Transfer Act and the Truth in Lending Act, respectively, Moore (2006). impose some insurance requirements. FRBNY Economic Policy Review / September 2008 143 effective without some system of penalties for noncompliance. anticipated and therefore adequate safeguards and procedures Monetary fines serve as deterrents. Contracts and laws assign may not be in place to address them. Emerging methods face a legal liability for failures, which can be costly if breached, while learning curve when confronting these issues. As evidenced by some regulations establish performance standards and impose their survival and success, established payment methods have penalties when they are not met. Varian’s (2004) theoretical devised ways to mitigate these risks. The key question regarding analysis of the polder model, described earlier, suggests that emerging payment methods is whether their providers have the relatively severe penalties—beyond the economic cost of a incentives and means to overcome the risks that could security lapse—may be necessary to ensure compliance. otherwise hinder widespread adoption. Limitations on liability mean that penalties cannot do the Competition gives important incentives to payment whole job. In cases of fraud, the party most deserving of method providers to mitigate many of these risks. Users can punishment, the fraudster, is usually long gone by the time the choose from many payment methods, and their choices reflect fraud is discovered. Even in cases where liability for a fraud or the extent to which the methods best facilitate smooth, low-risk transactions. In competition with payment methods less susceptible to fraud or operational failures, providers of new [There are] a variety of techniques— payment methods have clear incentives to address those risks. Failure to do so jeopardizes a method’s viability. As in other pricing, insurance, and containment—for markets, competition among payment methods is an creating incentives for participants in retail important mechanism to induce providers to address these payment transactions to mitigate fraud problems. New payment technologies can improve economic welfare and operational risks. by allowing diverse participants—consumers, merchants, banks, and nonbank service providers—to exchange payment data in ways not previously possible. The value of these operational incident can unambiguously be assigned to a technologies hinges, of course, on data integrity. Successful known party, there may be no practical level of penalty that payments systems will find ways of coordinating the behavior could cause the guilty party to internalize the consequences of diverse parties to facilitate data exchanges that serve their of its inadequate risk controls. Sometimes the threat of the mutual best interests. ultimate penalty—exclusion—may be the most effective deterrent: Payments system participants that fail to maintain adequate operational standards or fraud controls may be barred or expelled from the system. Data Integrity and Privacy Thus, we have a variety of techniques—pricing, insurance, and containment—for creating incentives for participants in Integrity of payment data is important not only as a safeguard retail payment transactions to mitigate fraud and operational against fraud and operational interruptions, but also for risks. Underlying structural aspects of many electronic retail maintaining participants’ privacy. Privacy issues have come to payments—particularly their nonrival nature—and the the fore in recent months. A group known as the Privacy Rights concomitant ability to limit access to the payment networks Clearinghouse reports that more than 165 million records have make containment techniques especially useful for creating been compromised by data security leaks since February deterrence tools. 2005.20 Such data breaches create potential for fraud, identity theft, and general loss of privacy. Similar to other aspects of payment data integrity, the maintenance of participants’ privacy constitutes a nonrival 3.2 Special Concerns for Emerging good. By preserving the privacy of its legitimate participants, Payments Systems a payments system encourages widespread participation and enhances the value of the system to all users. But as discussed Any viable payments system must find ways to maintain the above, nonrivalness can make it difficult to reach agreement integrity of payment data, but certain concerns are unique to among payments system participants on the necessary level of emerging payment methods. privacy protection. First, there is a “newness factor.” The novelty of emerging payment methods implies that various problems may not be 20 See 144 Understanding Risk Management in Emerging Retail Payments Maintaining privacy is tricky because, by nature, it runs value can be transferred. While the relative speed of the counter to the payment function: Every type of payment transactions is generally a desirable feature in the general requires the exchange of some information, which under the market—it reduces certain types of fraud—it can also make it wrong circumstances can be subject to misuse. For a consumer difficult to identify and preempt illicit transactions. to use a credit card to buy something from a merchant, for Similarly, some emerging payment methods involve highly example, he or she must give the credit card information to transportable stores of value, either in physical or electronic the merchant. The consumer’s surrender of credit card form. Diverse participation and a high degree of privacy, both information is essentially a compromise between the of which are features that make a payments system attractive merchant’s need to identify the consumer and the consumer’s for legitimate users, can make it easier to mask illicit use. Some desire to remain anonymous to prevent misuse of his or her personal information. The merchant obtains enough information about the consumer to determine that the Rather than creating financial risk to direct transaction is legitimate, but no more. Under some circum- participants in a [payment] transaction, stances, maintaining privacy can conflict with the goal of preventing fraud, as Stigler (1980) points out. Moreover, illicit use introduces or carries broader Katz and Hermalin (2006) discuss efficiency reasons for societal risks. privacy that suggest that the full sharing of private information within a payments system could be inefficient, even in the presence of fraud risk. Every successful payments system has emerging payment methods operate with little or no involve- to reach a workable compromise between these two facets ment of conventional financial intermediaries such as banks, of transaction privacy. making it difficult for authorities to monitor and identify illicit use. Network connectivity addresses the breadth of uses of a payment method and may alter its attractiveness as a store of value. The interfaces through which transactions are initiated Illicit Use may alter the ability to identify illicit transactions. In practice, it may be hard to distinguish between “user-friendly” and Unlike many risks associated with payments systems, the use “illicit-user-friendly” platforms. of a payment method for illicit purposes (such as money Like other types of nonrival goods, payments systems can laundering, financing of terrorism or crime, or the purchase guard against illicit use through the use of monitoring and of illegal goods) rarely involves direct risk of financial loss to penalties (including criminal penalties) and through the a participant in the payment transaction. Thus, unlike many exclusion of miscreants. But the high degree of similarity operational risks, the use of a payment method for illegal between the needs of legitimate and illegitimate users of activities does not pose a risk as such for other users of that payment technologies, as well as the need to balance societal payment method. In this case, the payment method works costs and benefits, suggests that some amount of criminal use as designed, but individuals use the method for nefarious and other socially undesirable activity will always slip through. purposes external to the payments system itself. Rather than Society’s determination of what constitutes an acceptable creating financial risk to direct participants in a transaction, threshold of illicit use is a complex and thorny issue that goes illicit use introduces or carries broader societal risks. Since beyond the scope of this article. monetary gains are one determinant of the level of criminal activity, erecting obstacles and deterrents to these activities supports an important public good. Unfortunately, many of the features that provide value for 4. Three Examples of Risk and legitimate transactors can also make them susceptible to Its Management in Emerging misuse by individuals engaging in money laundering and Payments terrorism financing. Features that suggest the potential for a payment method to be used or misused for illicit purposes We present three informal case studies to illustrate how include speed of value transfer, transportability, inter- characteristics of new payment methods affect potential risk, mediation, anonymity, quantity limits, network connectivity, how key participants act to mitigate those risks, and how and ease of interface. A common feature of many of these participants’ actions demonstrate the economic principles methods, especially electronic methods, is the speed with which described above. The three payment methods are: 1) general- FRBNY Economic Policy Review / September 2008 145 purpose prepaid cards, 2) e-check payments through the ACH Risk Analysis system, and 3) proprietary online balance-transfer systems such as PayPal. Each incorporates new technologies, new The advance-payment feature substantially mitigates credit or networks, and new rules to create an entirely new payment nonpayment risk in general-purpose prepaid card products, method. These examples are not intended to demonstrate the allowing such cards to be marketed widely and distributed full range of payment options. They are used in different directly to consumers by nonbank third parties, referred to as venues, employ different means for initiating payments, and card sponsors. Although every payment card must be issued by clear and settle transactions differently; yet they employ similar a bank, a nonbank sponsor’s logo often appears as the most risk mitigation strategies. (The appendix describes our prominent brand name on the card. The broad involvement of selection of the case studies.) nonbank institutions in the distribution of general-purpose While the payment methods in these case studies are not prepaid cards stands in contrast to the common practices of immune to all types of risk, we concentrate on fraud, traditional debit and credit card programs. operational, and illicit use risks because emerging methods Since general-purpose prepaid cards use the credit and debit appear particularly susceptible to these problems. The case card infrastructure for transactions, clearing, and settlement, studies focus on those areas of emerging payments that differ they share the risks inherent in these more mature financial from established payment types. To the extent that an products. These cards also exhibit a number of new risks, emerging payment is initiated using new technology but clears including a complex supply chain that often involves nonbank and settles through an established settlement network, our third parties at vulnerable stages of delivery and an increased discussion examines the new front-end mechanism but susceptibility to money laundering and illicit transactions. excludes the clearing and settlement portion. For general-purpose prepaid cards, nonbank institutions often stand between the cardholder and the bank that issues the card. In many cases, the nonbank institutions maintain the 4.1 General-Purpose Prepaid Cards primary relationship with the cardholder. This prominent role for a third party in initiating and maintaining customer General-purpose prepaid cards, branded by a payment network such as Visa, MasterCard, American Express, or Discover, can be used by all merchants that accept that network Since general-purpose prepaid cards use brand. Introduced in the 1990s, the cards function similarly to the credit and debit card infrastructure for credit and debit cards at a point of sale: A customer swipes a plastic card through a standard reader, and the transaction is transactions, clearing, and settlement, authorized and settled through a card network. In addition, they share the risks inherent in these more some cards can be used to withdraw funds from ATMs and to make remote purchases or pay bills, similar to debit cards.21 mature financial products. Cardholders can often check the balances available on their cards through a website or telephone response system. General-purpose prepaid card programs differ in price, relationships can complicate the regulatory treatment of cards product functionality, customer identification requirements, and introduce credit risk for the bank issuers and, potentially, value limits, and levels of cardholder protection. Their the cardholders. The third-party entities could go bankrupt or distinguishing characteristic is that they require cardholders be subject to various operational failings that would be less to turn over funds in advance for future purchases of goods likely to impact accounts at a supervised and FDIC-insured and services. Frequently, the funds on these cards can be financial institution. The involvement of the major card reloaded at a variety of outlets, such as at merchants, over the associations and the fraud detection that they bring to bear Internet, or through ATMs. This feature allows a cardholder appear to deter illegal activity. News reports recount instances to use a single card without replacement or interruption, thus of fraud, however, such as using stolen credit cards to purchase increasing the card’s value as a potential substitute or prepaid cards at a self-serve checkout counter.23 complement to a formal banking relationship.22 22 See McGrath (2007) for further discussion of the functionality and market 21 Payroll cards are a similar application, but differ dramatically in terms of the position of prepaid cards. business model used for marketing and distribution and in terms of regulatory 23 David Hench, “Savvy Thieves Use Gift-Card Scam to Fool the System,” coverage. Payroll card programs are not discussed here. Portland Press Herald/Maine Sunday Telegram, February 21, 2007. 146 Understanding Risk Management in Emerging Retail Payments Additionally, third-party nonbanks may not have the same laundering.27 These include capping the stored dollar amount level of data security that banks have, potentially exposing per card, limiting the frequency with which and the value of consumer data to greater risk of theft. In particular, third-party funds that can be reloaded, obtaining and confirming certain distributors may fail to impose uniform data security standards customer data prior to approving card applications, and for their retailers, a security lapse that increases risks for data providing liability protection for consumers in the event of gathered and stored at the point of sale. card loss or fraudulent usage. A downside of the flexibility provided by cards able to The operational and fraud risks of general-purpose prepaid facilitate nearly anonymous transactions is that they are cards are evidenced by: 1) a more complicated supply chain for attractive vehicles for abuse by illegal enterprises.24 The Drug Enforcement Administration, Immigration and Customs Enforcement, and Internal Revenue Service–Criminal The major card networks have issued Investigation each allege that prepaid cards are used in bulk guidelines to the issuing banks that are cash smuggling. They explain that drug dealers load cash onto prepaid cards and send them to their drug suppliers outside the intended to reduce the attractiveness of country who use the cards to withdraw money from a local ATM. prepaid cards for money laundering. This potential for illicit use is exacerbated if card issuers or sponsors operate offshore because it makes it harder to enforce relevant regulatory requirements. In fact, some general- providing the cards, often involving nonbank third parties in purpose prepaid card products are openly marketed as a primary customer relationships, and 2) the potential for illicit convenient way to circumvent law enforcement and tax transfer of funds. Domestic and international law enforcement authorities. For example, a prepaid card called the Freedom officials are particularly interested in mitigating the latter risk. Card used to promise, among other things, “a fully anonymous ATM debit card . . . requiring no phone numbers or IDs . . . no daily cash withdrawal or loading limits . . . real- 4.2 ACH e-Checks time card funding with any e-currency, PayPal, Western Union or bank wires.”25 The card was originated by an offshore Over the last decade, the National Automated Clearing House financial institution, but it could be used to obtain funds Association (NACHA), which sets rules for ACH transactions, throughout the world. This product appears not to exist any has gradually developed rules and formats for six new more. While such offerings are often short-lived, dubious new electronic debit transactions, referred to as e-checks. These products emerge regularly. e-checks allow banks and their clients to convert checks or information from checks into ACH debits.28 The following discussion describes two types of e-check transactions— Mitigation accounts-receivable conversion (ARC) and telephone-initiated (TEL) transactions—to illustrate the risk factors and Efforts are also under way to deal with the illicit use and data mitigation trade-offs associated with these new transaction security risks. An industry task force says it is in the process types. of creating “AML [anti-money laundering] best practices ARC rules permit businesses to transform checks mailed by guidelines” in response to anticipated regulations from the U.S. bill-paying consumers into ACH debits. In the fourth quarter Treasury’s Financial Crimes Enforcement Network aimed at of 2006, the 613 million ARC transactions initiated accounted thwarting money laundering and terrorist financing through 27 As reported in Money Laundering Threat Assessment Working Group (2005). 26 prepaid cards. The major card networks have issued 28 NACHA members include financial institutions and regional clearinghouse guidelines to the issuing banks that are intended associations. NACHA manages the development, administration, and to reduce the attractiveness of prepaid cards for money governance of the ACH system. Its rules provide more than fifteen worktype codes for different types of payments—such as corporate-to-corporate 24 payments, recurring payments, point-of-sale payments, and e-check See Money Laundering Threat Assessment Working Group (2005), Financial payments—as a means to identify specific rules, formats, and uses. Historically, Action Task Force (2006), and Sienkiewicz (2007). the ACH has typically been used for direct deposit of payroll and Social Security 25 See FRBNY Economic Policy Review / September 2008 147 Return Rates for Selected Automated Clearinghouse (ACH) Applications, 2002-07 Percentage of volume Percentage of volume 1.6 4.0 Unauthorized Returns Administrative Returns 1.4 3.5 1.2 3.0 1.0 TEL 2.5 TEL 0.8 2.0 0.6 1.5 0.4 1.0 ARC ARC 0.2 0.5 PPD PPD 0 0 2002 03 04 05 06 07 2002 03 04 05 06 07 Source: Federal Reserve System. Notes: Return rates are calculated from monthly Federal Reserve ACH data; ARC, PPD, and TEL are accounts-receivable conversion, preauthorized payment and deposit, and telephone-initiated transactions, respectively. for more than half of e-check volume. Use grew by about specific return codes indicate high levels of specific risks. one-third over the prior year: Roughly 6 percent of checks For example, various processing-problem codes identify written are now being converted to electronic debits under administrative returns that can indicate operational problems, ARC rules.29 whereas returns of transactions not authorized by an account TEL transactions are debits to consumers’ accounts holder (known as unauthorized transactions) can indicate authorized by the account holder via telephone to a merchant, fraud problems. vendor, or service provider. These transactions make one-time The chart presents historical rates for ARC and TEL for the ACH payments available when written authorizations are not types of returns most likely to suggest administrative or fraud feasible.30 The 76 million TEL transactions processed during problems. It also shows parallel return rates for preauthorized the fourth quarter of 2006 reflect a 16 percent increase over the payments and deposits (called PPD), which is the most widely previous year. TEL transactions account for about 7 percent of used recurring, non-e-check debit transaction and serves as a e-check volume. useful comparison. ACH transactions that fail to clear (because of, for example, There are similarities and differences in the records of ARC insufficient funds, errors in processing, or suspected fraud) and TEL returns. In the fourth quarter of 2002, ARC return are returned along with a code indicating the reason for the rates for both unauthorized and administrative returns were return. Return rates are useful indicators of risk because similar to those for PPD debits, but by the fourth quarter of the standardized return reason codes indicate what type of 2006 they had fallen below the rates for PPD debits. In the problem caused the clearing failure. Typically, high levels of fourth quarter of 2002, TEL return rates were some six to eight times the rates for PPD debits, and although they have since 29 Bank for International Settlements (2006, p. 157) reports that 33.1 billion fallen, TEL return rates remain at least twice those for PPD checks were paid in 2005. ARC transactions are written as checks but paid as debits. Overall, ARC and TEL each had high return rates in electronic debits. their introductory periods and the rates declined over time. 30 NACHA rules restrict TEL transactions to prevent their use for “cold-call” telemarketing, but they can be used when there is a preexisting relationship ARC return rates today are very low, whereas those for TEL between merchant and consumer or when the consumer initiates the call. remain relatively high. 148 Understanding Risk Management in Emerging Retail Payments Risk Analysis businesses, including some telemarketers and bill collectors, that may not have received sufficient scrutiny or monitoring ARC and TEL transactions share some risks with other ACH from the banks through which they originate their debit transactions, but differ in risks that are driven by the transactions. Inadequately researched bank relationships can location at which the payments are initiated and the relation- undermine the “gatekeeper” function in this payment method, ships among the parties to each transaction.31 For ARC, retail making it difficult to deny dishonest originators access to the lockbox processors convert checks sent by consumers to billers. ACH network. Use of third-party service providers for TEL can Lockbox staff use high-speed equipment to capture coded compound the difficulty of identifying illegitimate initiators by information from the remittance slips and checks. The lockbox adding an intermediary between the payment-originating bank business is highly concentrated, mature, and controlled. In and its ACH debit-originating clients. many cases, these processors operate as subcontractors to the originating banks, supporting the banks’ cash management product offerings. In contrast, TEL transactions rely on customer input of account information via telephone, a Mitigation context in which the data and customer’s identity cannot easily be verified. Many of NACHA’s rules and procedures aim to control and For ARC, the largest risk is operational. ARC rules initially mitigate these risks. NACHA defines the rights and made business checks ineligible for conversion to an ACH responsibilities of ACH participants, including originators debit. Early problems included inadvertent conversion of (merchants, lockbox operators, and other businesses that business checks, particularly those that are the same size as initiate ACH payments) and originating banks (banks that provide ACH services to originators). Originators are required to follow NACHA rules and procedures when preparing and Inadequately researched bank submitting ACH payments. Originating banks warrant certain relationships can undermine the aspects of ACH transactions and are financially liable for returned transactions. To help control this liability, originating “gatekeeper” function in [the telephone- banks typically use contract language to shift risk to initiated] payments system. originators. Originators are thus given financial incentives to correct and avoid processing problems. When the problem arose of business checks being converted consumer checks. Banks on which these checks were drawn inappropriately, ARC originators reconfigured processing often returned the transaction.32 Another early problem was equipment to improve separation of business checks and that processors could not properly match the ARC payment to worked to change NACHA rules to permit conversion of the the appropriate checking account: During a pilot program, one business checks that were hardest to identify.34 Originators also bank reported its associated administrative returns were as reduced administrative returns by building databases to match 33 high as 10 percent. ARC payments and checking accounts.35 NACHA requires a For TEL transactions, fraud is a larger risk, perhaps lockbox processor or its bank to keep check images for two augmented by operational risk caused by various participants years, but to destroy the physical check within fourteen days. in its supply chain. TEL is designed for ad hoc transactions Such measures help mitigate fraud risk and simultaneously between merchants and consumers, some of whom do not have decrease the risk of processing a check twice. a preexisting relationship. A long-standing business NACHA rules also require originating banks to gather relationship is thus often absent, which increases the likelihood sufficient information to understand the background and of either seller or buyer fraud. Adding to this risk is the fact that business of any new originator that may be given access to ACH TEL opened the ACH network to new merchants and services. This gatekeeping function generally keeps dishonest 31 Shared infrastructure and processes can contribute to risks in certain ACH originators out of the ACH network, but it proved inadequate payments. The ACH network does not use real-time authentication and for TEL transactions. As illustrated by the Assail example authorization, and there are no centralized databases of originators accused of fraudulent use of the ACH system. These are mitigation techniques used by 34 Daniel Wolfe, “Dealing with the Accidental Conversions,” American Banker, other payment networks. December 8, 2004. Note that effective September 15, 2006, business checks that 32 Daniel Wolfe, “Dealing with the Accidental Conversions,” American Banker, do not carry an indicator in the auxiliary on-us field of the MICR line can be December 8, 2004. converted to ACH debits. 33 Steve Bills, “Pilot Done, Wells to Widen Lockbox Conversion Effort,” 35 Steve Bills, “Pilot Done, Wells to Widen Lockbox Conversion Effort,” American Banker, October 18, 2002. American Banker, October 18, 2002. FRBNY Economic Policy Review / September 2008 149 described earlier, telemarketing was one source of fraud that 4.3 Proprietary Online Balance-Transfer resulted in high return rates for TEL transactions. Systems As evidence of problems with TEL transactions mounted, NACHA intervened directly with originating banks and Among the payment options that arose for Internet commerce outside of its normal processes. NACHA and its member banks are proprietary online schemes to transfer balances of funds identified the specific TEL originators responsible for initiating between accounts. In this type of scheme, customers establish many of the transactions that were subsequently returned, and an account with a service provider, such as PayPal, and use 36 these originators were shut down. e-mail messages to initiate payments.41 If both parties to a Various participants in the ACH system have taken steps payment have accounts with the same service provider, the to improve the effectiveness of originating banks as service provider simply transfers monetary balances between gatekeepers. These steps include the introduction of NACHA their accounts. At PayPal, most customers are buyers and rules requiring originating banks to screen and monitor sellers (small businesses and individuals) involved in online 37 originators and to execute appropriate contracts. transactions, usually at an auction site. The service is also used Additionally, in June 2003, NACHA instituted a monitoring by small online companies and by individual customers who process to flag originators with TEL returns exceeding 2.5 per- value the ability to transfer funds from person to person. cent. Outside of the NACHA framework, ACH operators have Neteller, a similar service provider, is widely used for payments introduced risk monitoring services and rule changes, and federal regulatory agencies increased the attention given to these transactions in their guidelines on controlling risk in Among the payment options that arose for retail payments.38 These actions were followed by a rapid decline in returns, suggesting that monitoring, enforcing Internet commerce are proprietary online rules, and limiting access to the ACH network have been schemes to transfer balances of funds successful strategies for risk mitigation. between accounts. Although these mitigation efforts reduced TEL return rates, the return rates remain higher than NACHA would like. Thus, NACHA is pursuing additional proposals to make monitoring to online gaming sites.42 Other online person-to-person return items and resolving problems more effective. To address payment providers that follow a proprietary balance-transfer risk issues beyond those of TEL more broadly, NACHA also or similar model include GreenZap, StormPay, and eGold.43 reorganized its risk management infrastructure, creating a Risk We call these providers proprietary online balance-transfer Management Advisory Group to help implement a new risk systems. 39 management framework. Subsidiary work groups are PayPal, the largest and most well-known online payment attempting to address three areas of risk mitigation: 1) control service provider, uses the proprietary online balance-transfer of access to the ACH system, 2) the monitoring and control approach and intermediated almost $23 billion in transactions environment, and 3) enforcement activity. Additionally, to during the first half of 2007. PayPal is larger and more increase the visibility of risk management at ACH originating sophisticated than any of its competitors. eBay, the huge online banks, the Office of the Comptroller of the Currency issued a auction business, acquired PayPal in 2002, and eBay guidance document in September 2006 requiring that key ACH statistics be reported to banks’ boards of directors and senior 40 See OCC Bulletin no. 2006-39, “Automated Clearing House Activities: Risk officials.40 Management Guidance,” available at 150 Understanding Risk Management in Emerging Retail Payments transactions currently generate almost 70 percent of PayPal’s the seller does not fulfill on commitments. GreenZap also states dollar volume. that members send funds to third parties at their own risk.46 These service providers act as agents by accepting deposits Online businesses are also vulnerable to the risk of outages, from customers and allowing money to be transferred from and businesses with high visibility seem to be most attractive to one in-house account to another. Although specific those seeking to disrupt services and overcome security arrangements vary, in-house account balances typically are features. The size of PayPal (about 133 million accounts as funded from a bank account by ACH transfer or a buyer’s/ of year-end 2006) and the speed at which technical changes sender’s credit or signature debit card. Frequently, funds can be withdrawn by check or by co-branded debit/ATM card, transferred to the user’s individual bank account by ACH Proprietary online balance-transfer credit, or used for future transfers within the network. payment methods depend on complex, multistep processes. For the user, the tasks are kept simple. Behind the scenes, Risk Analysis however, many parties . . . are involved Although the volume of activity suggests that this type of in completing a transaction. payment meets a market demand for rapid online payments, it remains an emerging payment method accompanied by a variety of risks. Examples of fraud, operational, and illicit use are made to support its growth have, indeed, led to some risks include: 1) fraud associated with simple enrollment and significant system downtime and made PayPal subject to anonymity; 2) operational errors and malicious attacks, such as hacker attacks.47 In addition, in October 2004, a site redesign 44 “phishing” and “pharming”; 3) operational risk associated crippled some of its operations, leaving the website unavailable with technological complexity and a complex supply chain; for two days and subject to intermittent outages for several days and 4) susceptibility to illicit use. Specific rules, processes, thereafter.48 Moreover, PayPal and eBay were the top phishing controls, and screening capabilities vary across providers, targets in 2005, representing 62 percent of all attacks, according yielding different levels of unmitigated risk and affecting the to Netcraft, a company that tracks and blocks phishing sites.49 availability of mitigation options. Proprietary online balance-transfer payment methods The core philosophy of the proprietary online balance- depend on complex, multistep processes. For the user, the tasks transfer model is to permit easy, quick entry and 24-hour are kept simple. Behind the scenes, however, many parties availability. Under this system, an unknown, possibly (including individuals, merchants, third-party service anonymous, seller can be positioned to perpetrate fraud or providers, the buyer’s and seller’s banks, and the ACH, debit simply fail to live up to his or her side of a transaction. Such a card, and credit card networks) are involved in completing a dishonest seller could take the money and not ship the product. transaction. As is the case generally in complex networks, the The buyer would then have to try to recover funds under the large number of digital “hands” and handoffs increases the rules of the payment provider’s user agreement or protection difficulty of identifying and assessing risk severity and the policy. To be covered under PayPal’s Buyer Protection Policy, exposures that can vary by user, channel, or product. the seller must enroll in the verification program and the buyer Intentional user anonymity makes these services susceptible must comply with other eligibility requirements.45 In contrast, to illicit use, such as money laundering or payments for illicit the user agreement for GreenZap, a smaller but similar service purposes. Only the service provider has information about user provider, indicates that it is not liable for any purchases or identities. While this structure protects users from fraud and services and does not issue refunds for a product or service if 46 GreenZap claimed 777,600 users at year-end 2006. See FRBNY Economic Policy Review / September 2008 151 identity theft, it can also make it easier for users to transfer behavior. While these efforts have not totally eliminated fraud, funds illegally because traditional enforcement authorities do they appear to have had some success: Statistics reported in the not have identifying information. In addition, theft of press show that merchants using PayPal have loss rates due to identities outside of the network could provide criminals with fraud that are noticeably below the e-commerce average.52 sufficient information to set up false accounts that can be used To prevent the risk of a data breach, PayPal says that it for illegal funds transfers. Further, if a service provider allows collects, encrypts, and stores sensitive customer information international transfers, the payment process might be used to on servers not connected to the Internet. Additionally, to launder funds between domestic and offshore accounts. counter phishing and pharming, PayPal provides clear instructions on what to do if customers suspect they have received a phishing e-mail. When notified of a phishing attack, Mitigation PayPal attempts to close down the perpetrator’s site within twenty-four hours.53 As the leading service provider, PayPal has an incentive to PayPal limits its own risk inherent in its complex supply invest in good risk management tools and oversight to protect chain by specifying its own rights and responsibilities as well as its payment method. Its risk management, in turn, protects those of its users in cases where errors, disruptions, or legitimate users and establishes standards for other online unauthorized transactions occur. The user agreement is payment service providers. The following examples illustrate complex, and it is updated as needed. Information on how PayPal establishes contracts or manages relationships with its suppliers is not publicly available. Online payment service providers have Online payment service providers have addressed the risk of illicit use and international exposure by placing limits on addressed the risk of illicit use and transfers and account balances for unverified accounts.54 international exposure by placing limits PayPal relaxes these limits for its verified accounts, but the on transfers and account balances for verification process exposes would-be criminals. As a result, PayPal may have become less useful for money laundering. It unverified accounts. does appear possible, however, to launder large sums of money by sending small increments to many accounts using a mass- pay type of function.55 To counter the above risks, eBay and that PayPal, in conjunction with eBay, appears to have learned PayPal have established a joint fraud investigation team to from its losses and risk exposures, creating systems, track down problem transactions and users. Moreover, within technologies, and rules that help control the risks that emerged the context of its legal obligations, PayPal has a strong history in its early years. As a result of its efforts, PayPal says that its loss of cooperating with law enforcement agencies.56 rate is four-tenths of 1 percent, well below that of the credit 50 Ultimately, the proprietary online balance-transfer model is card industry. a self-contained, closed payment method, albeit one open to a To combat machine-based attacks, PayPal developed an wide range of potential participants. All payment account account creation process that requires manual human input, activity occurs within a single entity, which can make it easier which has blocked unmanned computer “bots” from opening for a service provider to internalize and control risks. By accounts. It also created multiple levels of service, in which operating as a closed system, a service provider can manage higher levels of account service require additional identity confirmation. The verified member program, for example, 52 Paul Cox, “PayPal and FBI Team Up,” Wall Street Journal, June 22, 2001; protects PayPal and creates a product it markets to customers. Ralph F. Wilson, “Assessing Criticism of PayPal,” Web Commerce Today, PayPal also retains the right to terminate service to any March 15, 2002; Rob Garver, “eBay and Banking: Is PayPal a Serious Rival?” participant it suspects of not complying with its rules.51 American Banker, November 15, 2005. 53 Similarly, Neteller’s annual report describes significant expansion in its In addition, PayPal developed background computer fraud, security, and IT capabilities. See 152 Understanding Risk Management in Emerging Retail Payments fraud by denying or restricting access to users who do not meet CardSystems, and TJX incidents, even interruptions of low- its membership eligibility requirements or who fail to provide value payments can result in large losses and disruption of the required authentication. It can temporarily or permanently business for many participants. block users who do not comply with its rules or who are suspected of fraudulent or unauthorized activities. PayPal’s experiences illustrate that a provider must aggressively battle new operational and fraud threats with vigilant monitoring 5.2 Maintain a Perimeter of payment transactions. All legitimate payments system participants—consumers, merchants, banks, and other service providers—share a common interest in risk mitigation. The nonrival nature of risk mitigation means that all these participants operate behind the 5. Lessons Learned same common protective perimeter of security and reliability. Successful payment methods find ways to encourage an The foregoing case studies offer many useful lessons for managing the problems that arise in emerging payment appropriate buy-in of all participants in terms of contributing methods. Although each case is unique, there are common to this shared resource. As the case studies illustrate, themes, which can be organized into three basic lessons. wrongdoers need to be kept outside this perimeter—even in the most inclusive payment methods. PayPal offers a good illustration of this principle. A key aspect of PayPal’s market positioning is its openness, 5.1 Recognize the Problem inclusiveness, and ease of use. It claims that all anyone needs to participate in PayPal is an e-mail address. However, as PayPal The very features that contribute to the efficiency of new has become more sophisticated and has placed increased value payment forms—their scalability, speed, and relative on avoiding fraud and operational losses, it has accordingly anonymity—can also enable the rapid proliferation of various tightened its perimeter and imposed participation standards. types of payment risk. As information moves more easily among payments system participants, more intensive management is needed to safeguard this data flow. Moreover, All legitimate payments system the more widespread and successful the system becomes, the participants—consumers, merchants, bigger the potential for disruptions. The incident reported earlier concerning two Russian men banks, and other service providers—share scamming PayPal offers a striking illustration of this principle. a common interest in risk mitigation. The perpetrators first breached the security arrangements at Internet service providers, gaining an initial cover of anonymity. They then used electronic means to create Today, PayPal screens each participant, requiring not only an anonymous e-mail accounts, which in turn were used to create e-mail address but also some identifying information as well as bogus accounts at PayPal. The speed and extent to which this credit card, debit card, or bank account information (all of was possible relied fundamentally on computers and the which can be independently verified) before a participant is Internet. permitted to send funds. To date, most innovative payment methods still have Telephone-initiated ACH transactions offer another relatively low volumes of transactions. So even if risks are example of adaptation to new risks. The highly decentralized not well controlled, the overall risk of loss is limited.57 nature of the ACH, in which debit transactions are created by a Complacency, however, would be irresponsible. Significant wide variety of entities, has facilitated a relatively high fraud flaws or fraud risks to ACH products have the potential to rate in the case of TEL transactions. Recent and proposed reach more institutions and individuals than most emerging changes to NACHA rules are meant to encourage buy-in from payment products. And, as demonstrated by the Assail, banks in controlling this problem. They do so by imposing monitoring of problematic originators and, under some 57 The volume and value of e-money payment transactions in the United States proposed rules, penalties for violators. This process is are negligible. In contrast, ACH e-check transactions grew more than 40 per- cent last year, totaling about 2 billion transactions for the first half of 2006. necessarily more complicated than it is in a proprietary system See Bank for International Settlements (2006, Tables 7 and 8 and pp. 145-6). such as PayPal, given the diverse composition of the ACH. Yet FRBNY Economic Policy Review / September 2008 153 there now seems to be widespread acceptance of the idea that zation along the payment supply chain represents a source of stringent rules—such as exclusion—are required to keep fraud efficiency, but the heavy involvement of nonbank or third- rates down to manageable levels.58 party participants means that the defensive perimeter for data Prepaid cards are something of an intermediate case, being integrity cannot be monitored by the banking system alone. neither purely proprietary like PayPal nor as decentralized as Historically, the role of third-party processors was limited the ACH. On the one hand, anyone can purchase a prepaid to back-office services, such as lockboxes. In conjunction with card at a retail outlet and anyone, not necessarily the same emerging payment methods, some third-party entities have individual, can make a purchase with the card at a participating moved into the more prominent position of maintaining primary relationships with customers. Conversely, in some cases, banks have moved from maintaining primary relation- For prepaid cards, the card associations ships to becoming back-office service providers. This role serve as enforcers to ensure the integrity reversal for bank and nonbank institutions has raised policy concerns and is a topic that warrants additional study.62 of the network, for example, by minimizing operational and fraud risks. 5.3 Trust the Marketplace—but Not Blindly retailer. On the other hand, the card association whose name is Producing a nonrival good is always a difficult and often a on the card (for example, MasterCard or Visa) screens issuing controversial business. Computer software, recorded music, banks and binds them contractually to particular provisions. and video, three common examples, are frequent objects of The card associations also screen the card-selling merchants for public controversy, regulation, and litigation. But somehow, a variety of risks, including the effectiveness of their security to the market finds innovative ways to provide these goods prevent large-scale theft. Finally, the card associations impose fairly—though rarely without growing pains along the way. contractual and monitoring provisions on merchants that Electronic payment services also demonstrate both market- 59 accept their cards. driven discipline and creativity, including for their security and For prepaid cards, the card associations serve as enforcers to reliability components. New payment products are ensure the integrity of the network, for example, by minimizing immediately subjected to the forces of a market’s “invisible operational and fraud risks. Thus far, this control appears hand,” including ramifications of exposure to operational, to be effective, even though the nominal issuers of general- fraud, and data security risks. As a result, operators are forced purpose prepaid cards—merchants and various third parties— to learn about previously undetected operational problems. are neither typical card issuers nor regulated financial Outages of almost any sort can rapidly undermine user institutions. In a broader context, the aftermath of the May confidence in the reliability of a product, a particular service 2005 data breach at CardSystems illustrates the efficacy of such provider, or a new form of payment generally. New products control:60 Visa and American Express subsequently barred also seem to attract the attention of fraudsters eager to exploit CardSystems from participating in their networks, forcing the flaws before they are rectified. Only if a payment provider can firm out of business.61 address such problems quickly and effectively can it stay in The CardSystems case also highlights the difficulties posed business. Thus, for many of these risks, market mechanisms by lengthening supply chains in the payment industry. Again, provide significant incentives for service providers to see that tensions can arise between efficiency and security. Speciali- they are addressed promptly and thoroughly. New products in their early stages repeatedly show 58 NACHA has recently approved a code of conduct that establishes standards patterns of operational or fraud problems and unmitigated of behavior and “specifies NACHA’s right to disassociate itself from any risk, after which containment efforts follow. When PayPal organization that, in NACHA’s opinion, fails to meet the standards and principles stated in the code” (emphasis added). See Elliott C. McEntee, faced fraud losses early on, it took steps to reduce those losses. “Open Letter,” NACHA, April 13, 2006. It also implemented new authentication techniques and 59 See, for example, BankInfoSecurity.com, “Visa Takes Aim at Data introduced innovative technology. PayPal continues to revise Companies,” August 8, 2006. 60 See Perry (2005). 62 Concerns about the role of nonbank third parties in the payments system 61 CardSystems was purchased by Pay by Touch for its merchant network, have been raised, but they remain unresolved. See, for instance, Hoenig (2000) according to a company press release dated October 15, 2005. and Sullivan (2007). 154 Understanding Risk Management in Emerging Retail Payments its contracts and participation agreements to increase general market incentives. The federal government and many controls and limit risk. states respond to this risk by enacting laws and regulations to Some providers of similar online payment services have prohibit the use of payment methods for such purposes and by failed, at least in part because of fraud losses, and others have creating incentives for payment providers to screen out run into trouble with law enforcement authorities over illicit prohibited transactions. A measured policy response again payments.63 The service providers that survive are those that seems appropriate, as the risk of illicit use must be balanced are able to identify and mitigate losses quickly. When NACHA against the costs of compliance. introduced the TEL product in 2001 and return rates began to soar, it took steps to identify the source of the problems. As a result, return rates fell to more acceptable levels. The WEB transaction, another recently created ACH e-check application 6. Conclusion useful for Internet transactions, had a return rate of 0.68 per- cent in 2002, but it fell to 0.08 percent in 2004.64 Innovative payment mechanisms, such as the ones described in As payments systems grow and flourish, however, so too this article, are making transactions less expensive and easier, does the potential for disruption. Recent developments in the while opening new commercial venues for payment payment card industry provide an illustration. Card networks, transactions. As with more established forms of payment, historically quite vigilant in the protection of their data however, the ultimate success of these inventive arrangements will depend on their ability to control risk. For retail payments, the predominant risks are operational, As payments systems grow and fraud, illicit use, and data security risks. Providers mitigate flourish . . . so too does the potential these risks through techniques such as pricing, insurance, and containment. In the growing market of electronic transactions, for disruption. these techniques have shared value that does not decline with additional use and can be enhanced with additional contributions—in other words, they are nonrival. integrity, have nonetheless been subject to significant data This article examined three emerging payment methods to breaches. Increasing volume and a more diffuse supply chain draw some lessons from their operation and markets. The have posed new difficulties. The card networks have responded payment methods explored here carry transactions that are by putting more pressure on merchants to comply with data relatively low in value, and, during their start-up phases, most security standards, but this effort remains a work in progress.65 had a small number of users. However, some ACH-based The vitality of the market for payment services does not rule transactions quickly reached substantial volume levels. With out a role for public policy. Well-designed regulations can help low values and generally limited breadth, the payment methods coordinate industry efforts and maintain industry standards. do not currently pose systemic risks or demonstrate substantial Laws and criminal penalties can serve as deterrents to activities policy gaps. We note, however, that the risks discussed here are such as fraud. In addition, the importance of confidence in the not confined to emerging payments. overall payments system—a public good—should not be All payment processes have risks that must be controlled. underestimated. Policymakers will always have an interest in Fraudsters seem especially drawn to new technologies, ensuring that disruptions in one method of payment, however becoming early adopters in their attempts to exploit any unlikely, do not spill over into other segments of the payments identifiable weaknesses. But fraudsters can also perpetrate system. innovative attacks against established systems. Moreover, even In contrast to other risks considered here, the steps needed low-value retail payment providers can be the targets of to reduce the risk of illicit use are not always fully supported by machine-based attacks that can cause substantial damage; the speed of corruption and potential for proliferation of 63 See, for example, Neteller Lawrence Complaint: United States of America v. damaging problems are certainly shared by all payment Stephen Eric Lawrence, Southern District of New York. January 16, 2007, available at FRBNY Economic Policy Review / September 2008 155 considering risk and mitigation strategies that can be used to the network’s safeguards. The applicability of this approach analyze new as well as established payment methods. to the risk of illicit use, however, appears less certain. Our analysis of the risk mitigation techniques used by More cooperative, open systems, which derive some of payments system providers concludes that containment is the their utility from their universality, have less ability to exclude dominant means of controlling risk. Generally, market particular users and thus face greater risk mitigation mechanisms appear to encourage providers to mitigate risks challenges. Nonetheless, the problems, risks, and gaps in appropriately: Most private-sector providers have the tools to processes can be addressed only if the providers and manage many of these risks, particularly because they treat the participants remain vigilant while applying the lessons we integrity of the network as a club good; in other words, they described. retain the option to exclude any party that fails to comply with 156 Understanding Risk Management in Emerging Retail Payments Appendix: How the Case Studies Were Selected In general, new payment methods are built on top of existing • General-purpose prepaid cards use established card- products. Enhancements, significant innovations, and various swipe technology to create a new payment and therefore levels of rule changes are added to these products to address fall in the upper-right quadrant. newly identified market opportunities or to take advantage of • Accounts-receivable conversion (ARC) uses the new expanding technological capabilities. To identify the extent to access channel of scanning technology and software to which a payment method is new rather than more established, read paper checks and create transactions that flow over we grouped the components of a generic payment process into the established ACH network, as represented in the lower- two broad categories: the access channel and the payment left quadrant. method.a An access channel is used at the beginning of the transaction process; it provides the user interface or front end • Proprietary balance-transfer systems meld a new access (for example, a plastic card with a magnetic stripe) and may or technology—the Internet—with new transaction may not include verification of the identity of the involved methods—e-mail and balance transfers—and therefore parties and validation of the payment instrument. The payment fall into the lower-right quadrant. method includes the remaining parts of the payment process The TEL and ARC payments are types of ACH e-checks that governed by applicable laws, regulations, and contracts. share a clearing and settlement network and many rules (these These various factors—new versus established components are addressed jointly in the analysis above). of access channels and payment methods—can be organized into a 2 x 2 matrix, as shown in Exhibit 1. Payment methods that have the fewest changes from established methods fit a See Bank for International Settlements (2000) for a description of the into the upper-left quadrant, although rule changes or new components of payment processes. combinations of established characteristics can yield a new payment method. The lower-right quadrant includes emerging payment methods that incorporate the greatest number of new Exhibit 2 characteristics in terms of both access channels and payment Examples of Payments methods. The remaining two quadrants, upper right and lower Noncash Payment Methods left, are hybrids of new and established components. Transaction, Clearing, and Settlement Processes Exhibit 2 provides examples of payments that might be found Established New in each of these four cells. For the case studies, we selected one payment method from each quadrant (shaded). Credit card payments General-purpose Paper checks in general prepaid cards • ACH payments initiated via telephone (TEL) fall in the hed s PPD ACH debit Closed network and upper-left quadrant, since neither the telephone access PPD ACH credit gift cards tabli channel nor the ACH clearing and settlement portions s PIN debit are new. E Signature debit s s ACH debit card hed Exhibit 1 s ACH TEL s tabli Micropayment aggregators Access Channels and Payment Methods s Check 21 substitute check component Noncash Payment Methods of e New combination Transaction, Clearing, and Settlement Processes Channel ss ACH POP Proprietary balance Established New End Front ACH ARC transfer Acce Secured proprietary Well-known technologies Existing access ACH WEB Check image presentation balance transfer hed initiate commonly known technologies initiate a s (not IRD) Proprietary balance s types of payments or use new type of payment. Cell-phone payment transfer via cell phone tabli new rules to create new s Highway toll booths Prepaid wallet E types of payments. New Contactless card payments Instant credit Channel (debit or credit) ss Charge to phone bill Front End Front New technologies or New technologies and Biometic authentication networks access networks initiate new Acce License ID to create New established payment types of payments. ACH debit method. ACH credit push FRBNY Economic Policy Review / September 2008 157 References Anderson, R., and T. Moore. 2006. “The Economics of Information Hoenig, T. M. 2000. “Payments and Settlement Systems: Future Security.” Science 314, no. 5799 (October 27): 610-3. Players and Issues.” Remarks delivered at the BAI Money Transfer 2000 Conference, Chicago, Illinois, November 9. Available at Bank for International Settlements. 2000. “Clearing and Settlement Hirshleifer, J. 1983. “From Weakest Link to Best Shot: The Voluntary Provision of Public Goods.” Public Choice 41, no. 3 (January): 371-86. 158 Understanding Risk Management in Emerging Retail Payments References (Continued) Stigler, G. J. 1980. “An Introduction to Privacy in Economics and Varian, H. R. 1998. “Markets for Information Goods.” Remarks Politics.” Journal of Legal Studies 9, no. 4 (December): 623-44. prepared for a Bank of Japan conference, June 18 and 19. Available at FRBNY Economic Policy Review / September 2008 159 Antoine Martin and David C. Mills An Economic Perspective on the Enforcement of Credit Arrangements: The Case of Daylight Overdrafts in Fedwire • The Federal Reserve’s extension of daylight 1. Introduction overdrafts to banks exposes the central bank to some credit risk during the day. redit arrangements between a borrower and a lender are a C prevalent part of the economy. A fundamental concern for • The Fed manages this exposure through a any lender is the risk that the borrower fails to fully repay the loan combination of tools, including monitoring, as expected, a type of risk called credit risk. Thus, lenders want an awareness of banks’ reputations, and credit arrangements that are designed to compensate them for— collateral requirements. and help them effectively manage—credit risk. In certain situations, central banks engage in credit • Under a proposed policy change, the Fed arrangements as lenders to banks. For example, the Federal would supply intraday balances to healthy Reserve offers certain banks overnight loans at the discount banks through collateralized and window. Additionally, it provides liquidity to many banks uncollateralized overdrafts; banks would be during the day whenever those banks must overdraw on their allowed to pledge collateral voluntarily to Federal Reserve accounts in order to make payments and settle support intraday overdrafts. securities. This extension of daylight overdrafts by the Fed can be interpreted as very-short-term credit, so the central bank is exposed to credit risk that it must manage. • An analysis of the increased use of collateral This article discusses how the Federal Reserve manages its resulting from the change points to potential credit risk exposure from daylight overdrafts. We first present benefits—as well as costs—for the Federal a simple economic framework for thinking about the causes of Reserve, banks, and the financial system. credit risk and the possible tools that lenders have to help them Antoine Martin is a research officer at the Federal Reserve Bank of New York; The views expressed are those of the authors and do not necessarily reflect the David C. Mills is an economist at the Board of Governors of the Federal position of the Board of Governors of the Federal Reserve System, the Federal Reserve System. Reserve Bank of New York, or the Federal Reserve System. FRBNY Economic Policy Review / September 2008 161 manage it. We then apply this framework to the Federal consequences of the original decision (once the loan is Reserve’s Payments System Risk Policy, which uses a variety of obtained because of the promise to repay, the borrower no tools to manage credit risk. Finally, we discuss a possible increase longer wants to repay it). Anticipating this outcome, the lender in the use of collateral as a credit risk management tool, as may choose to forgo making the loan in the first place. presented in a recent policy proposal published by the Board of Governors of the Federal Reserve System (hereafter, the Board) that considers changes to its Payments System Risk Policy (Board of Governors of the Federal Reserve System 2008). 2.2 Enforcement To achieve a good outcome, borrowers would like to be able to credibly commit to not reneging on their promise. A strong 2. A Framework for Thinking about enough commitment can sufficiently address the time- inconsistency problem.1 Experience shows, however, that this Contractual Relationships kind of commitment is difficult to make. Institutions, formal Economists have developed a framework for thinking about or informal, that help economic agents make credible contracts in general and credit arrangements in particular. We commitments are said to provide enforcement. Courts are an now summarize and illustrate the main elements of this example of such institutions, but many other examples exist. If enforcement were costless to lenders, they could framework. The emphasis is on first principles, an approach adequately control opportunistic behavior and it would not that provides a helpful basis for policy analysis. affect the decision to lend and the determination of the interest rate to charge. Lenders would typically charge an interest rate sufficient for them to cover the risk of bad luck in the credit 2.1 Bad Luck versus Opportunistic Behavior relationship. A borrower may not fully repay a lender for one of two reasons: bad luck or opportunistic behavior. By “luck,” we mean all 2.3 Information Problems random factors that affect borrowers’ and lenders’ actions and that are independent of their behavior. For example, weather is Enforcement is rarely, if ever, costless. In particular, a lender a random factor that can influence a farmer’s yield of corn may have inadequate information about some actions or traits independent of the amount of effort the farmer exerts. The of the borrower. Economists distinguish between two types of effect of luck can typically be priced into a contract. information problems: moral hazard (or hidden actions) of the In contrast, opportunistic behavior—a privately beneficial borrower and adverse selection (or hidden types) of borrowers.2 action that increases costs to the other party in the transaction— Consider this example of moral hazard. It is well understood typically cannot be priced into a contract. Opportunistic behavior that if a bicycle is insured against theft, its owner is less likely to occurs when borrowers may not have sufficient incentive to do all protect it as carefully as if it were not insured. The hidden they can to repay their debts. In the example of the farmer, the action here is how carefully the owner protects the bicycle. lender wants the farmer to put forth great effort to yield a large Since the insurance company is unable to observe this action, it amount of corn and would like to be assured that the farmer will cannot make the insurance contract dependent on it. do so. The farmer is opportunistic if he does not work very hard in An example of the adverse selection problem is found in the the field. By not working very hard, he may not yield enough corn health insurance industry. The hidden type here is an to fully repay his debt to the lender. Why do borrowers have an incentive to engage in 1 A classical example of solving a time-inconsistency problem with credible opportunistic behavior? At the time a credit arrangement is commitment is found in the Greek myth of Ulysses and the sirens. Ulysses made, all borrowers promise to repay their debt. Otherwise, would like to hear the song of the sirens, but knows that once he does, he will be compelled to change the course of his ship and crash it against the rocks on lenders would refuse to lend. Once the loan is made, however, which the sirens are standing. To enforce his commitment to not change borrowers have an incentive to renege on their promise and course, Ulysses asks his sailors to bind him to the mast of the ship and to put default. The economic decision of the borrower is time wax in their ears. The wax will prevent the sailors from hearing the sirens and from hearing Ulysses when he asks his sailors to change course. Such elegant inconsistent. In other words, the best decision at a given time solutions to an enforcement problem are, unfortunately, not always available. (the promise made at the beginning of the credit arrangement 2 Moral hazard and adverse selection are terms from the insurance markets, to repay the loan) may no longer be optimal later because of the where these problems were first studied. 162 An Economic Perspective individual’s health. Someone who seeks health insurance which is both observable and verifiable. Economists therefore knows his or her health better than the insurance company classify contractual situations as either of two types: 1) those in does. Individuals who believe they are likely to need a lot of which perfect enforcement is possible because all relevant medical attention, for instance, will want to choose insurance information is publicly available and 2) those in which there is with better coverage. only imperfect or costly enforcement because at least some In the context of a lending relationship, the lender may not relevant information is private. be able to observe what the borrower does with the loaned funds. In the farming example, will the farmer buy equipment that will allow for a greater yield of corn, increasing the likelihood that the loan will be repaid? Or will the farmer 2.4 Enforcing Contracts through Reputation, instead buy a big-screen television, leading to lower effort and Monitoring, and Collateral making it less likely that the loan will be repaid? Similarly, the lender may not be able to observe the borrower’s type. A lender The information frictions described above can create credit risk will be more reluctant to lend to a farmer who has previously over and above the risk that might come only from bad luck. defaulted on other loans, which would suggest that this Because of information frictions, a more sophisticated policy borrower is not a good type compared with a farmer who has than simply charging an interest rate for a loan might be never defaulted. necessary. As we observed, when credit risk arises only from Borrower actions or types may be hidden from the lender, in bad luck, no additional policy is necessary because nothing can which case they are called unobservable. Alternatively, they may be done to affect the probability that the loan will be repaid. be observable to the lender but hidden to parties outside the In principle, there are several ways to alleviate enforcement lending relationship, such as courts of law, in which case they problems, and each method is costly. Because enforcement is are called unverifiable. In the previous example, the farmer’s not perfect, a trade-off always exists between better enforcement of contract terms and more costly means of ensuring enforcement. Among the ways of enforcing contracts Enforcement is rarely, if ever, costless. In in situations of imperfect information are reputation, particular, a lender may have inadequate monitoring, and collateral. We consider each of them in turn and provide an example of how a loan for a construction information about some actions or traits of project uses all three. the borrower. Economists distinguish between two types of information problems: moral hazard (or hidden Reputation actions) of the borrower and adverse In cases of repeated interactions, the terms of a contract can depend on past actions. Borrowers can obtain better terms selection (or hidden types) of borrowers. by establishing a reputation for good behavior. Reputation is achieved by showing a willingness to refrain from short- term opportunism. Reputation can be thought of as a way to use of fertilizer is observable and verifiable if the lender is able make private information about one’s type more public. In to ascertain whether fertilizer was used and if, in addition, a particular, it signals to a potential lender that a borrower is court is able to establish that fact. However, the quality of the more interested in long-term outcomes (possibly because he fertilizer used may be observable but not verifiable if, for or she wants to avoid punishments that restrict access to example, the lender can analyze the fertilizer but cannot prove future loans) than any short-term gains achieved by to a court, or any other third party, that the farmer used a defaulting on a loan. Reputation, therefore, can typically particular fertilizer of a certain quality. Finally, whether the alleviate problems associated with adverse selection. In the farmer used the correct amount of fertilizer may be case of a construction loan, a building contractor who wants unobservable and unverifiable because neither the lender nor a to finance a new project may rely on reputation in third party can determine how much fertilizer was used. negotiating terms for a new loan. A solid credit history Information that is either unobservable or unverifiable is increases the contractor’s chances of securing a new loan typically called private—as opposed to public information, and allows him to negotiate favorable terms. FRBNY Economic Policy Review / September 2008 163 Monitoring This conversion can provide the borrower with more favorable terms, such as a lower interest rate. Lenders can prevent opportunism by closely monitoring Depending on the circumstances, some of these ways to borrowers’ actions, by screening and certifying their quality alleviate enforcement problems may be more or less costly or and that of their project, or, after a default, by verifying the efficient. Reputations may be costly or impossible to maintain quality and amount of their assets and operations. Monitoring if there are not enough opportunities to signal one’s type—for can be thought of as a way to acquire information that would example, if relationships are short lived or if the economic otherwise be private. As a result, monitoring can typically help environment evolves quickly and in unpredictable ways. alleviate the incentive issues associated with both moral hazard Monitoring can be difficult or costly because it may require and adverse selection. In monitoring construction loans, for very specific and technical knowledge or because it may be example, lenders conduct periodic inspections and require possible to misrepresent the true state of a project. The use of status reports from the contractor or independent third parties collateral, too, is not without cost; there are costs involved in as a way to keep track of the project’s progress. valuing and managing it. The collateral may have more value to the borrower than the lender, which implies that, in the case of default, the collateral is transferred from one agent that gives it Collateral a higher valuation to another agent that assigns it a lower valuation. This reallocation results in a loss to society. There By posting collateral, the borrower offers a type of guarantee to may also be a cost associated with rationing credit if the the lender. Collateral may be something that has value to the collateral is insufficient. lender so that the lender is at least partially compensated in case Finally, technological advances can also change the relative of default. In that particular case, the collateral plays an costs and benefits of the various ways of alleviating enforcement insurance role and need not have any value to the borrower. problems. For example, innovations in information technology Collateral may also be something that has value to the borrower have improved recordkeeping and the transmission of infor- so that its loss punishes the borrower in case of default. In that mation. The effect of improvements in information technology case, the collateral plays an incentive role and need not have has likely reduced the costs of reputation, monitoring, and any value to the lender. In practice, collateral typically plays collateral, making these tools more effective at reducing information problems. The ability to keep better records enables borrowers to signal information about their reputations. It also The effect of improvements in information allows lenders to gather and evaluate information quickly, which technology has likely reduced the costs of reduces the cost of monitoring. Furthermore, better information technology can improve lenders’ evaluations of certain assets that reputation, monitoring, and collateral, can be pledged as collateral, reducing some uncertainty regarding making these tools more effective at the collateral’s value. reducing information problems. 3. The Case of Daylight Overdrafts both roles in that it usually has some value to both the borrower on Federal Reserve Accounts and the lender. It is the incentive role that is most important from the perspective of reducing information frictions. Thus, We now turn to the specific case of the Federal Reserve’s policy collateral typically helps alleviate the incentive problems regarding daylight overdrafts on accounts that banks have at associated with moral hazard. the Fed.3 Most, but not all, of the value of overdrafts arises from Various assets can be pledged as collateral. For example, banks’ Fedwire activity.4 Fedwire is a large-value payments loans for such durable goods as houses, cars, and boats are system and a securities settlement system that banks use to send often secured by the goods themselves. In the financial sector, securities and other financial assets can be used as collateral for 3 See Board of Governors of the Federal Reserve System (2007). various types of loans. In our construction loan example, once 4 Overdrafts can also arise from check clearing and settlement via the a project is complete and a building is ready for sale, the Automated Clearing House services provided by the Federal Reserve. The Federal Reserve’s overdraft policy applies to the net account balance resulting contractor can convert the loan into a standard mortgage, from activity over all Federal Reserve services to banks. Here, we focus on which requires that the new building be pledged as collateral. Fedwire because most of the value of overdrafts is the result of Fedwire activity. 164 An Economic Perspective a each other funds and government securities on behalf of their Frictions in the Payments System customers and their own accounts. Transactions are sent over Search Friction Fedwire one at a time with finality, which means that the A search friction refers to the efforts that would be necessary for Federal Reserve guarantees that the funds or securities a bank a payer (the party that intends to send a payment to some other 5 receives will not be revoked. Because transactions are party) and potential liquidity providers to make contact with one processed one at a time, banks must have access to enough another and to determine the right amounts of liquidity to transfer funds to complete each transaction. This need for available to the payer’s accounts. If a payer did not have sufficient funds in funds generates various frictions that banks face in the its account and did not have access to overdrafts provided by the settlement of transactions, such as search frictions, timing central bank, it would have to borrow the amount of the payment frictions, and incentive frictions (see box). prior to sending it. But from whom should it borrow? The payer The Federal Reserve alleviates the impact of these frictions would not necessarily know which other party has sufficient funds 6 by providing intraday liquidity, which allows qualifying banks in its account, and so it must search for such a lender. to overdraw on their Fed accounts in order to make payments Timing Friction via Fedwire. Banks can acquire overdrafts throughout the day The timing friction refers to the operational difficulty of achieving to make payments, but must ensure that their accounts are not the precise timing for when funds will be delivered during the day. in a negative position at the end of the day. The Federal Even if parties overcome the search friction and agree on a Reserve’s provision of liquidity through daylight overdrafts can particular amount of funds to be delivered by one participant to be interpreted as very-short-term credit. another for the purpose of funding some time-critical payments of This exposure is something the Federal Reserve must the borrower, how will the borrower be assured of receiving the manage to protect itself from moral hazard or adverse selection funds at the given time? A commercial bank may have operational problems that may arise from the type of information frictions difficulties or experience delays for other reasons. The borrower described earlier. For example, because the Fed does not would simply have to wait for delivery of the funds, which reduces observe all the actions of banks, it may be concerned that some the benefits of the arrangement. banks could use daylight overdrafts to finance excessively risky bets. Similarly, the Reserve Banks may not have full Incentive Frictions There are two incentive frictions to confront in adapting to a information regarding a bank’s risk of default on daylight withdrawal of daylight credit. First, the rewards of providing overdrafts. The Fed currently manages its exposure to this form intraday funding need to appropriately reflect the costs and risks of of credit risk with a combination of overdraft fees, reputation, doing so. This is also true with overnight funding arrangements, monitoring, and collateral. We now turn to some specifics of but the intraday timing possibly exacerbates these frictions. the policy to make this connection clearer. Lending $1 billion overnight at a 4 percent interest rate yields The Federal Reserve charges an explicit price for daylight approximately $111,000 in earnings, but lending it for an hour at overdrafts, currently a twenty-four-hour rate of 36 basis points the same rate would yield only $4,600. Assuming the processing less a deductible. This price, though small, is meant to provide costs for arranging the delivery and return of funds are fixed and an incentive for banks to minimize their use of daylight roughly similar for an intraday and an overnight loan, then it may overdrafts. But even though this fee may help constrain the size not be profitable for potential lenders to enter the market at low of daylight overdrafts, and accordingly the Federal Reserve’s interest rates. credit exposure, it does not address the information frictions of Second, payments system participants have the option to delay adverse selection and moral hazard. Thus, other aspects of the sending a payment rather than borrowing, if the cost of borrowing policy address those issues. is too high. However, if all participants are inclined to delay, the The daylight overdraft fee provides some incentive for system may be vulnerable to gridlock. banks to constrain the size of their daylight overdrafts. In addition, the Fed uses a net debit cap, which is the maximum a dollar amount of daylight overdrafts that an institution may This material is borrowed from McAndrews (2006). incur in its Federal Reserve account. Each bank that has an 5 Although transactions cannot be revoked, that does not mean that they cannot be reversed. Reversals, however, are conducted by initiating a second account and that is also eligible for intraday overdrafts has a net irrevocable transaction. debit cap. The policy on net debit caps is based on a set of 6 Note that some of these frictions are attributable to imperfect information and the absence of commitment. However, we focus here on the incentive specific guidelines and some degree of banking supervision. problems arising from the provision of intraday liquidity by the Reserve Banks. The policy allows for one of six ratings for a bank. For most FRBNY Economic Policy Review / September 2008 165 banks, net debit caps range from zero to 2.25 times the bank’s The Federal Reserve’s policy regarding daylight overdrafts risk-based capital.7 uses a combination of fees, monitoring, reputation, and Net debit caps involve a great deal of monitoring. The collateral. Changes in payments and securities settlement Federal Reserve reviews supervisory information, evaluates systems, and their effect on the need for intraday liquidity, have banks’ self-assessments (if applicable), and then uses this led to periodic reviews of this policy to determine whether information to assess the appropriateness of an institution’s changes to it can improve the safety and efficiency of the cap category.8 This monitoring alleviates some problems payments system. Recently, in order to ease intraday liquidity associated with adverse selection. constraints and reduce operational risk, the Board proposed The Federal Reserve also monitors a bank’s use of its changes to its Payments System Risk policy to supply intraday daylight overdrafts against the cap, providing an opportunity balances to healthy banks predominantly through explicitly for banks to establish reputations with their regional Federal collateralized daylight overdrafts.10 Under the proposal, the Reserve Bank. In most instances, banks that exceed their cap Board would allow banks to voluntarily pledge collateral to limit are required to explain the reason to the Fed and then be support intraday overdrafts. Collateralized intraday overdrafts counseled to prevent it from happening again. The Fed reserves would be charged a zero fee, while the fee for uncollateralized overdrafts would increase from 36 to 50 basis points.11 We now describe how increasing the use of collateral could bring The Federal Reserve’s policy regarding benefits as well as costs to the Federal Reserve, to banks, and to daylight overdrafts uses a combination the financial system as a whole. of fees, monitoring, reputation, and collateral. 3.1 The Benefits of Increasing Collateral Use the right to reduce net debit caps unilaterally, impose Greater use of collateral has the potential to benefit the Federal collateralization or clearing-balance requirements, reject or Reserve, banks, and the financial system in several ways. delay certain payments, or, in extreme circumstances, prohibit The Federal Reserve could benefit because collateral the bank from using Fedwire. Thus, maintaining a reputation provides it with some insurance in the event a bank cannot of staying under the cap can help banks avoid such actions and repay its overdraft. It may also benefit if greater use of collateral can alleviate certain moral hazard concerns. increases the incentives for banks to repay their overdrafts over Although most daylight overdrafts are uncollateralized, and above the incentives already in place because of the Federal Reserve uses collateral in two situations. First, it monitoring, reputation, and the existing use of collateral. requires collateral from problem institutions to cover any Banks could benefit if greater use of collateral relaxed some incidental overdrafts. Second, banks wishing to increase their credit constraints. As we observed in the construction loan net debit caps can pledge collateral to do so subject to the example, providing collateral can often allow a borrower to Federal Reserve’s approval. The amount and type of collateral obtain better terms on a loan. For similar reasons, the Board’s pledged are determined through an agreement between the policy proposal includes a zero fee on collateralized daylight bank and the Federal Reserve.9 Collateral plays an insurance overdrafts. In such a case, banks’ overdraft costs could decrease role for the Fed in the event of a loss due to an overdraft. It also when they pledge collateral. plays an incentive role for the bank to control its overdrafts and The financial system may benefit if the increased use of avoid risky behavior that could lead to its closure and forfeiture collateralized intraday overdrafts at the zero fee speeds up the of the assets it pledged as collateral. Thus, the collateral here flow of payments across financial markets. The lower cost of also overcomes certain moral hazard concerns. collateralized intraday overdrafts may lead to more payments being made earlier in the day, as banks would have less need to 7 For foreign banking organizations, a net debit cap is a function of no more delay payments until they have sufficient incoming funds. By than 35 percent of their worldwide capital (referred to as their U.S. capital equivalency). encouraging more banks to have collateral pledged at the Fed, 8 Each bank that uses a relatively large amount of overdrafts must perform a increased use of collateral could make it easier for the Federal self-assessment of its own creditworthiness, intraday funds management and control, customer credit policies and controls, and operating controls and 10 See Board of Governors of the Federal Reserve System (2008). contingency procedures. 11 There are other proposed changes as well. See Table 1 in Board of Governors 9 The type and value of collateral pledged are consistent with the Federal of the Federal Reserve System (2008) for a summary of all the proposed Reserve’s discount window policy. changes. 166 An Economic Perspective Reserve to inject liquidity both intraday and overnight in times could lead to credit rationing in the economy should a shortage of financial stress. This is true in particular because collateral is of collateral occur. Thus, the increased use of collateral could required for overnight loans. In addition, increased use of have negative spillover effects on the economy. Again, whether collateral may prepare banks for financial stress by increasing such a cost is likely to be large is an empirical question and their ability to borrow at the discount window. All of these depends on the range of collateral that would be acceptable to spillover benefits may accrue to the financial system through the Federal Reserve and other banks. greater use of collateral. As with any policy proposal, a careful analysis of the overall benefits and costs of a change in the daylight overdraft policy is essential. It should be noted that the costs of a greater use of collateral are higher for banks if collateral were required for all 3.2 The Potential Costs of Greater overdrafts. But if banks are given the choice between Collateral Use uncollateralized lending and posting collateral (with a zero fee on collateralized daylight overdrafts), then their costs should There are also possible costs to increasing the use of collateral. be lower because collateral would be pledged by those banks for The Federal Reserve could face higher costs associated with which it is the less costly option. monitoring collateral, such as making sure it is available and valuing it properly. The Federal Reserve already pays such costs because it accepts collateral for overnight loans, but these costs could rise if the amount of collateral increases. Moreover, if 4. Conclusion there is a greater reliance on collateral for intraday overdrafts, banks may ask to manage their collateral more actively at the How to best enable the extension of liquidity by a central bank is Federal Reserve, requiring the Fed to invest in enhancements to an important policy question. As the examples presented here its collateral management systems. suggest, it may be desirable to use a combination of reputation, Banks would have to pay costs associated with acquiring, monitoring, and collateral. However, the relative role of each managing, and tracking their collateral. Additionally, they may method of enforcing credit arrangements should depend on the face an opportunity cost associated with using collateral to details of the contractual relationship considered. secure overdrafts because that collateral may no longer be used In the future, we can expect the risk faced by central banks to for other purposes. Banks may also reallocate their portfolio of change over time, but we can also expect central banks to have assets to acquire enough collateral for daylight overdraft access to more effective enforcement technologies. As banks find purposes. Whether this would constrain banks much depends themselves in situations requiring them to take quick actions, on the type of collateral that the Federal Reserve and other credit risk can emerge unpredictably and without warning. banks are willing to accept.12 However, the quality of the tools used by central banks to mitigate The financial system as well may be negatively affected by the these risks has increased as well. For example, technological greater use of collateral. Collateralized overdrafts make the progress has the potential to make monitoring less costly and Federal Reserve a higher claimant on assets of a failed bank, more effective in the future. Moreover, new technologies could which reduces the attachable assets to residual claimants in the reduce risk in a number of ways: event of a bank liquidation, adversely affecting the unsecured • The development of liquidity-saving methods for safely creditors of that bank. This is an issue mainly if the policy is not transferring balances could reduce the demand for explained well in advance so that some long-term contracts daylight overdrafts. cannot be renegotiated. Another potential cost would occur if • The development of improved markets during the day too much of the banking system’s assets are tied to could potentially lead to a decreased demand for collateralized daylight overdrafts. In extreme situations, this intraday overdrafts, as they are replaced by better methods of intraday distribution of liquidity. In other 12 For discount window purposes, the Federal Reserve accepts a large range of words, the frictions that require the provision of daylight assets of varying liquidity and credit risk, by which they are categorized. The overdrafts today may be reduced by enhanced collateral value of each asset is a discounted value of an asset’s determined price. This applied discount is based on an asset’s class. If the discounts technology. accurately reflect a liquidity and risk premium, banks can have some flexibility • Technological progress will influence the need to rely in pledging collateral at the Federal Reserve and can minimize opportunity costs associated with collateral. The proposed policy would follow discount heavily on reputation, monitoring, or collateral as some window practices to determine which assets are acceptable, those assets’ of these methods may become relatively more effective categories, and the discount applied to the assets’ determined price. than others. FRBNY Economic Policy Review / September 2008 167 References Board of Governors of the Federal Reserve System. 2007. “Federal McAndrews, J. 2006. “Alternative Arrangements for the Distribution Reserve Policy on Payments System Risk” (as amended effective of Intraday Liquidity.” Federal Reserve Bank of New York January 11). Available at ———. 2008. “Policy on Payments System Risk.” Federal Register 73, no. 46 (March 7): 12417-43. 168 An Economic Perspective FEDERAL RESERVE BANK OF NEW YORK ECONOMIC POLICY REVIEW Single-copy subscriptions to the Economic Policy Review (ISSN 1932-0426) are free. All articles can also be downloaded at no charge from the Research Group’s website. Economic Policy Review articles may be reproduced for educational or training purposes, provided they are reprinted in full; include credit to the author(s), the publication, and the Bank; and include the publication’s disclaimer. Library of Congress Card Number: 77-646559 Federal Reserve Bank of New York research publications—the Economic Policy Review, Current Issues in Economics and Finance, Staff Reports, Research Update, and the annual publications catalogue—are available at the Research Group’s website. www.newyorkfed.org/research